In this paper we proposed a system; Optical Braille Translator (OBT), that identify Sinhala Braille characters in single sided Braille document and translates to Sinhala language. This system also capable of identifying Grade1 English Braille characters, numbers, capital letters and some words in Grade 2
English Braille system. Image processing techniques were used to developed the proposed system in MATLAB environment. The translated text displayed in a word application as the final outcome. Performance evaluation results reflect that the proposed method can recognize Braille characters and
translated to user selected language either Sinhala or English efficiently, over 99% of accuracy.
An Efficient Segmentation Technique for Machine Printed Devanagiri Script: Bo...iosrjce
Segmentation technique plays a major role in scripting the documents for extraction of various
features. Many researchers are doing various research works in this field to make the segmenting process
simple as well as efficient. In this paper a simple segmentation technique for both the line and word
segmentation of a script document has been proposed. The main objective of this technique is to recognize the
spaces that separate two text lines.For the Word segmentation technique also similar procedure is followed. In
this work ,three different scanned document have been taken as input images for both line and word
segmentation techniques. The results found were outstanding with average accuracy for both line and word. It
provides 100% accuracy for line segmentation and 100% for line segmentation as well. Evaluation results show
that our method outperforms several competing methods.
FREEMAN CODE BASED ONLINE HANDWRITTEN CHARACTER RECOGNITION FOR MALAYALAM USI...acijjournal
Handwritten character recognition is conversion of handwritten text to machine readable and editable form. Online character recognition deals with live conversion of characters. Malayalam is a language spoken by millions of people in the state of Kerala and the union territories of Lakshadweep and Pondicherry in India. It is written mostly in clockwise direction and consists of loops and curves. The method aims at training a simple neural network with three layers using backpropagation algorithm.
Freeman codes are used to represent each character as feature vector. These feature vectors act as inputs to the network during the training and testing phases of the neural network. The output is the character expressed in the Unicode format.
Automated Bangla sign language translation system for alphabets by means of M...TELKOMNIKA JOURNAL
Individuals with hearing and speaking impairment communicate using sign language. The movement of hand, body and expressions of face are the means by which the people, who are unable to hear and speak, can communicate. Bangla sign alphabets are formed with one or two hand movements. There are some features which differentiates the signs. To detect and recognize the signs, analyzing its shape and comparing its features is necessary. This paper aims to propose a model and build a computer systemthat can recognize Bangla Sign Lanugage alphabets and translate them to corresponding Bangla letters by means of deep convolutional neural network (CNN). CNN has been introduced in this model in form of a pre-trained model called “MobileNet” which produced an average accuracy of 95.71% in recognizing 36 Bangla Sign Language alphabets.
Two Methods for Recognition of Hand Written Farsi CharactersCSCJournals
This document describes two methods for recognizing handwritten Farsi characters using neural networks and machine learning techniques. The first method uses wavelet transforms to extract features from character borders and trains a neural network classifier on these features. It achieves 86.3% accuracy on test data. The second method divides characters into groups based on visual properties, extracts moment features for each group, and uses Bayesian classification with a decision tree post-processing step. It achieves an overall recognition rate of 90.64% according to the results presented. Experimental evaluations of both methods on different datasets of handwritten Farsi characters are discussed.
Handwriting character recognition (HCR) is the ability of a computer to receive and interpret handwritten input. Handwritten Character Recognition is one of the active and challenging research areas in the field of Pattern Recognition. Pattern recognition is a process that taking in raw data and making an action based on the category of the pattern. HCR is one of the well-known applications of pattern recognition. Handwriting recognition especially for Indian languages is still in infant stage because not much work has been done it. This paper discuss about an idea to recognize Kannada vowels using chain code features. Kannada is a South Indian language. For any recognition system, an important part is feature extraction. A proper feature extraction method can increase the recognition ratio. In this paper, a chain code based feature extraction method is investigated for developing HCR system. Chain code is working based on 4-neighborhood or 8–neighborhood methods. Chain code is a sequence of code directions of a character and connection to a starting point which is often used in image processing. In this paper, 8–neighborhood method has been implemented which allows generation of eight different codes for each character. These codes have been used as features of the character image, which have been later on used for training and testing for K-Nearest Neighbor (KNN) classifiers. The level of accuracy reached to 100%.
PERFORMANCE EVALUATION OF STATISTICAL CLASSIFIERS USING INDIAN SIGN LANGUAGE ...IJCSEA Journal
Sign language is the key for communication between deaf people. The significance of sign language is accentuated by various research activities and the technical aspects will definitely improve the communication needs. General view based sign language recognition systems extract manual parameters by a single camera view because it seems to be user friendly and hardware complexity; however it needs a high accuracy classifier for classification and recognition purpose. The decision making of the system in this work employs Indian sign language datasets and the performance evaluation of the system is compared by deploying the K-NN, Naïve Bayes and PNN classifiers. Classification using an instance-based classifiercan be a simple matter of locating the instance space and labelling the unknown instance with the same class label as that of the located (known) neighbour. Classifier always tries to improve the classification rate by pushing classifiers into an optimised structure. In each hand posture, a measure of properties like area, mean intensity, centroid, perimeter and diameter are taken; the classifier then uses these properties to determine the sign in different angles. They estimate the probability that a sign belongs to each of the target classes that is fixed. The impact of such study may reflect the exploration for using such algorithms
in other similar applications such as text classification and the development of automated systems.
A SIGNATURE BASED DRAVIDIAN SIGN LANGUAGE RECOGNITION BY SPARSE REPRESENTATIONijnlc
Sign language is a visual-gestural language used by deaf-dumb people for communication. As normal people are unfamiliar of sign language, the hearing-impaired people find it difficult to communicate with them. The communication gap between the normal and the deaf-dumb people can be bridged by means of Human–Computer Interaction. The objective of this paper is to convert the Dravidian (Tamil) sign language into text. The proposed method recognizes 12 vowels, 18 consonants and a special character “Aytham” of Tamil language by a vision based approach. In this work, the static images of the hand signs are obtained a web/digital camera. The hand region is segmented by a threshold applied to the hue channel of the input image. Then the region of interest (i.e. from wrist to fingers) is segmented using the reversed horizontal projection profile and the Discrete Cosine transformed signature is extracted from the boundary of hand sign. These features are invariant to translation, scale and rotation. Sparse representation classifier is incorporated to recognize 31 hand signs. The proposed method has attained a maximum recognition accuracy of 71% in a uniform background.
Recognition of Words in Tamil Script Using Neural NetworkIJERA Editor
In this paper, word recognition using neural network is proposed. Recognition process is started with the partitioning of document image into lines, words, and characters and then capturing the local features of segmented characters. After classifying the characters, the word image is transferred into unique code based on character code. This code ideally describes any form of word including word with mixed styles and different sizes. Sequence of character codes of the word form input pattern and word code is a target value of the pattern. Neural network is used to train the patterns of the words. Trained network is tested with word patterns and is recognized or unrecognized based on the network error value. Experiments have been conducted with a local database to evaluate the performance of the word recognizing system and obtained good accuracy. This method can be applied for any language word recognition system as the training is based on only unique code of the characters and words belonging to the language.
An Efficient Segmentation Technique for Machine Printed Devanagiri Script: Bo...iosrjce
Segmentation technique plays a major role in scripting the documents for extraction of various
features. Many researchers are doing various research works in this field to make the segmenting process
simple as well as efficient. In this paper a simple segmentation technique for both the line and word
segmentation of a script document has been proposed. The main objective of this technique is to recognize the
spaces that separate two text lines.For the Word segmentation technique also similar procedure is followed. In
this work ,three different scanned document have been taken as input images for both line and word
segmentation techniques. The results found were outstanding with average accuracy for both line and word. It
provides 100% accuracy for line segmentation and 100% for line segmentation as well. Evaluation results show
that our method outperforms several competing methods.
FREEMAN CODE BASED ONLINE HANDWRITTEN CHARACTER RECOGNITION FOR MALAYALAM USI...acijjournal
Handwritten character recognition is conversion of handwritten text to machine readable and editable form. Online character recognition deals with live conversion of characters. Malayalam is a language spoken by millions of people in the state of Kerala and the union territories of Lakshadweep and Pondicherry in India. It is written mostly in clockwise direction and consists of loops and curves. The method aims at training a simple neural network with three layers using backpropagation algorithm.
Freeman codes are used to represent each character as feature vector. These feature vectors act as inputs to the network during the training and testing phases of the neural network. The output is the character expressed in the Unicode format.
Automated Bangla sign language translation system for alphabets by means of M...TELKOMNIKA JOURNAL
Individuals with hearing and speaking impairment communicate using sign language. The movement of hand, body and expressions of face are the means by which the people, who are unable to hear and speak, can communicate. Bangla sign alphabets are formed with one or two hand movements. There are some features which differentiates the signs. To detect and recognize the signs, analyzing its shape and comparing its features is necessary. This paper aims to propose a model and build a computer systemthat can recognize Bangla Sign Lanugage alphabets and translate them to corresponding Bangla letters by means of deep convolutional neural network (CNN). CNN has been introduced in this model in form of a pre-trained model called “MobileNet” which produced an average accuracy of 95.71% in recognizing 36 Bangla Sign Language alphabets.
Two Methods for Recognition of Hand Written Farsi CharactersCSCJournals
This document describes two methods for recognizing handwritten Farsi characters using neural networks and machine learning techniques. The first method uses wavelet transforms to extract features from character borders and trains a neural network classifier on these features. It achieves 86.3% accuracy on test data. The second method divides characters into groups based on visual properties, extracts moment features for each group, and uses Bayesian classification with a decision tree post-processing step. It achieves an overall recognition rate of 90.64% according to the results presented. Experimental evaluations of both methods on different datasets of handwritten Farsi characters are discussed.
Handwriting character recognition (HCR) is the ability of a computer to receive and interpret handwritten input. Handwritten Character Recognition is one of the active and challenging research areas in the field of Pattern Recognition. Pattern recognition is a process that taking in raw data and making an action based on the category of the pattern. HCR is one of the well-known applications of pattern recognition. Handwriting recognition especially for Indian languages is still in infant stage because not much work has been done it. This paper discuss about an idea to recognize Kannada vowels using chain code features. Kannada is a South Indian language. For any recognition system, an important part is feature extraction. A proper feature extraction method can increase the recognition ratio. In this paper, a chain code based feature extraction method is investigated for developing HCR system. Chain code is working based on 4-neighborhood or 8–neighborhood methods. Chain code is a sequence of code directions of a character and connection to a starting point which is often used in image processing. In this paper, 8–neighborhood method has been implemented which allows generation of eight different codes for each character. These codes have been used as features of the character image, which have been later on used for training and testing for K-Nearest Neighbor (KNN) classifiers. The level of accuracy reached to 100%.
PERFORMANCE EVALUATION OF STATISTICAL CLASSIFIERS USING INDIAN SIGN LANGUAGE ...IJCSEA Journal
Sign language is the key for communication between deaf people. The significance of sign language is accentuated by various research activities and the technical aspects will definitely improve the communication needs. General view based sign language recognition systems extract manual parameters by a single camera view because it seems to be user friendly and hardware complexity; however it needs a high accuracy classifier for classification and recognition purpose. The decision making of the system in this work employs Indian sign language datasets and the performance evaluation of the system is compared by deploying the K-NN, Naïve Bayes and PNN classifiers. Classification using an instance-based classifiercan be a simple matter of locating the instance space and labelling the unknown instance with the same class label as that of the located (known) neighbour. Classifier always tries to improve the classification rate by pushing classifiers into an optimised structure. In each hand posture, a measure of properties like area, mean intensity, centroid, perimeter and diameter are taken; the classifier then uses these properties to determine the sign in different angles. They estimate the probability that a sign belongs to each of the target classes that is fixed. The impact of such study may reflect the exploration for using such algorithms
in other similar applications such as text classification and the development of automated systems.
A SIGNATURE BASED DRAVIDIAN SIGN LANGUAGE RECOGNITION BY SPARSE REPRESENTATIONijnlc
Sign language is a visual-gestural language used by deaf-dumb people for communication. As normal people are unfamiliar of sign language, the hearing-impaired people find it difficult to communicate with them. The communication gap between the normal and the deaf-dumb people can be bridged by means of Human–Computer Interaction. The objective of this paper is to convert the Dravidian (Tamil) sign language into text. The proposed method recognizes 12 vowels, 18 consonants and a special character “Aytham” of Tamil language by a vision based approach. In this work, the static images of the hand signs are obtained a web/digital camera. The hand region is segmented by a threshold applied to the hue channel of the input image. Then the region of interest (i.e. from wrist to fingers) is segmented using the reversed horizontal projection profile and the Discrete Cosine transformed signature is extracted from the boundary of hand sign. These features are invariant to translation, scale and rotation. Sparse representation classifier is incorporated to recognize 31 hand signs. The proposed method has attained a maximum recognition accuracy of 71% in a uniform background.
Recognition of Words in Tamil Script Using Neural NetworkIJERA Editor
In this paper, word recognition using neural network is proposed. Recognition process is started with the partitioning of document image into lines, words, and characters and then capturing the local features of segmented characters. After classifying the characters, the word image is transferred into unique code based on character code. This code ideally describes any form of word including word with mixed styles and different sizes. Sequence of character codes of the word form input pattern and word code is a target value of the pattern. Neural network is used to train the patterns of the words. Trained network is tested with word patterns and is recognized or unrecognized based on the network error value. Experiments have been conducted with a local database to evaluate the performance of the word recognizing system and obtained good accuracy. This method can be applied for any language word recognition system as the training is based on only unique code of the characters and words belonging to the language.
An Improved Approach for Word Ambiguity RemovalWaqas Tariq
Word ambiguity removal is a task of removing ambiguity from a word, i.e. correct sense of word is identified from ambiguous sentences. This paper describes a model that uses Part of Speech tagger and three categories for word sense disambiguation (WSD). Human Computer Interaction is very needful to improve interactions between users and computers. For this, the Supervised and Unsupervised methods are combined. The WSD algorithm is used to find the efficient and accurate sense of a word based on domain information. The accuracy of this work is evaluated with the aim of finding best suitable domain of word. Keywords: Human Computer Interaction, Supervised Training, Unsupervised Learning, Word Ambiguity, Word sense disambiguation
BrailleOCR: An Open Source Document to Braille Converter Applicationpijush15
This presentation is actually about an Open Source application, BrailleOCR that helps to convert scanned documents to Braille and thus helps the Visually Impaired.
What is the use of this application in real life? Well, BrailleOCR is currently the only app that integrated Optical character recognition and Braille Translation together. This app will eventually help converting a lot of important documents to Braille. The project site for this project is given here
IJCA Paper: http://www.ijcaonline.org/archives/volume68/number16/11664-7254
Project site: https://code.google.com/p/brailleocr/
The app uses a four step process. Initially, we have a scanned image, which is a RGB image. The first step or the Pre-Processing step deals with conversion of a RGB image to grayscale. The 2nd step deals with Character Recognition using the Tesseract Engine. Now, the recognition step may have errors and we require post processing to correct them. The 3rd step is thus the Post-Processing step and it actually corrects errors in the previous step. The final and the most important step is the Braille Conversion step.
Parameters Optimization for Improving ASR Performance in Adverse Real World N...Waqas Tariq
From the existing research it has been observed that many techniques and methodologies are available for performing every step of Automatic Speech Recognition (ASR) system, but the performance (Minimization of Word Error Recognition-WER and Maximization of Word Accuracy Rate- WAR) of the methodology is not dependent on the only technique applied in that method. The research work indicates that, performance mainly depends on the category of the noise, the level of the noise and the variable size of the window, frame, frame overlap etc is considered in the existing methods. The main aim of the work presented in this paper is to use variable size of parameters like window size, frame size and frame overlap percentage to observe the performance of algorithms for various categories of noise with different levels and also train the system for all size of parameters and category of real world noisy environment to improve the performance of the speech recognition system. This paper presents the results of Signal-to-Noise Ratio (SNR) and Accuracy test by applying variable size of parameters. It is observed that, it is really very hard to evaluate test results and decide parameter size for ASR performance improvement for its resultant optimization. Hence, this study further suggests the feasible and optimum parameter size using Fuzzy Inference System (FIS) for enhancing resultant accuracy in adverse real world noisy environmental conditions. This work will be helpful to give discriminative training of ubiquitous ASR system for better Human Computer Interaction (HCI). Keywords: ASR Performance, ASR Parameters Optimization, Multi-Environmental Training, Fuzzy Inference System for ASR, ubiquitous ASR system, Human Computer Interaction (HCI)
A bidirectional text transcription of braille for odia, hindi, telugu and eng...eSAT Journals
This document describes a bidirectional text transcription system for converting Braille documents in Odia, Hindi, Telugu, and English to text and vice versa using image processing techniques on an FPGA. It discusses prior work on Braille recognition and text-to-speech systems. The proposed algorithm segments Braille cells from image inputs and uses a modified database to map dot patterns to letters/words in each language based on number of dots. Letter patterns are rearranged for Hindi, Telugu, and English. The database is tested and dumped onto an FPGA for hardware implementation and bidirectional conversion between Braille and text in multiple languages.
IRJET - A Survey on Recognition of Strike-Out Texts in Handwritten DocumentsIRJET Journal
This document summarizes research on recognizing strike-out or crossed-out text in handwritten documents. It discusses 4 papers that studied this problem in different languages like English, Bengali, and Devanagari. The first paper used an LSTM model to recognize struck-out English text with 11% character error rate. The second used HMMs to recognize French text crossed out with lines or waves, achieving 46-92% accuracy. The third used graph algorithms and SVM to detect and remove Bengali strike-outs, obtaining 95% accuracy. The fourth used decision trees to classify English forensic documents, identifying 48% of crossed texts correctly. Overall the document reviews approaches to strike-out text recognition and removal.
Pattern Recognition of Japanese Alphabet Katakana Using Airy Zeta FunctionEditor IJCATR
Character recognition is one of common pattern recognition study. There are many object used in pattern recognition, such
as Japanese alphabet character, which is a very complex character compared to common Roman character. This research focus on
pattern recognition of Japanese character handwriting, Katakana. The pattern recognition process of a letter of the alphabet uses Airy
Zeta Function, with its input file is a .bmp file. User can write directly on an input device of the system. The testing of the system
examines 460 letter characters. The first testing that examines 230 characters result in an accuracy of 55,65%, whilst the second testing
that examines 460 characters produces an accuracy of 64,56% in recognizing the letters. These accuracy are much determined by the
quantity of training. The approach of pattern recognition is a statistical approach, where more pattern of letters are trained and saved as
a reference, more intelligent the system . The implementation of Airy zeta function methods in recognizing Japanese letter pattern is
able to produce high accuracy level.
The presentation will describe an algorithm through which one can recognize Devanagari Characters. Devanagari is the script in which Hindi is represented. This algorithm
could automatically segment character from the image of Devenagari text and then recognize them.
For extracting the individual characters from the image of Devanagari text, algorithm segmented the image several
times using the vertical and horizontal projection.
The algorithm starts with first segmenting the lines separately from the document by taking horizontal projection and then the line
into words by taking vertical projection of the line. Another step which is particular to the separation of
Devanagari characters was required and was done by first removing the header line by finding horizontal projection
of each word. The characters can then be extracted by vertical projection of the word without the header line.
Algorithm uses a Kohonen Neural Netowrk for the recognition task. After the separation of the characters from the
image, the image matrix was then downsampled to bring it down to a fixed size so as to make the recognition
size independent. The matrix can then be fed as input neurons to the Kohonen Neural Network and the winning neuron is
found which identifies the recognized the character. This information in Kohonen Neural Network was stored
earlier during the training phase of the neural network. For this, we first assigned random weights from input neurons
to output neurons and then for each training set, the winning neuron was calculated by finding the maximum
output produced by the neurons. The wights for this winning neuron were then adjusted so that it responds to this
pattern more strongly the next time.
The document discusses optical character recognition for Urdu handwriting. It introduces OCR and its applications. It then discusses earlier work on OCR systems that were font-specific. The document outlines the steps in OCR including image acquisition, preprocessing, segmentation, feature extraction, classification, and recognition. It provides an overview of the Urdu script and its variations. The document then summarizes research conducted on recognizing offline isolated Urdu characters using moment invariants and support vector machines. Other works discussed include an online and offline OCR system for Urdu using a segmentation-free approach, classifying Urdu ligatures using convolutional neural networks, and a segmentation-based approach for Urdu Nastaliq script recognition.
DEVNAGARI DOCUMENT SEGMENTATION USING HISTOGRAM APPROACHijcseit
This document summarizes a research paper on Devnagari document segmentation using a histogram approach. It discusses challenges in segmenting the Devnagari script used for several Indian languages. A simple algorithm is proposed using horizontal and vertical histograms to segment documents into lines, words and characters. The algorithm achieves near 100% accuracy for line segmentation but lower accuracy for word and character segmentation due to complexities in the Devnagari script. Future work is needed to improve character segmentation handling connected and modified characters.
This paper presents a new multi-tier holistic approach for recognizing Urdu text written in Nastaliq script. It first identifies special ligatures like dots, tay, hamza and mad from base ligatures. It then associates the special ligatures with neighboring base ligatures. Features are extracted from the ligatures and special ligature-base ligature associations. These features are input to a neural network that recognizes the ligatures in three steps: 1) identifying special ligatures, 2) associating them with base ligatures, and 3) recognizing the base ligatures. The system was tested on 200 ligatures with 100% accuracy for ligatures in its training set and closest match classification for new ligatures.
The Heuristic Extraction Algorithms for Freeman Chain Code of Handwritten Cha...Waqas Tariq
Handwriting character recognition (HCR) is the ability of a computer to receive and interpret handwritten input. In HCR, there are many representation schemes and one of them is Freeman chain code (FCC). Chain code is a sequence of code direction of a characters and connection to a starting point which is often used in image processing. The main problem in representing character using FCC that it is depends on the starting points. Unfortunately, the study about FCC extraction using one continuous route and to minimizing the length of chain code to FCC from a thinned binary image (TBI) have not been widely explored. To solve this problem, heuristic algorithms are proposed to extract the FCC that is correctly representing the characters. This paper proposes two heuristics algorithm that are based on randomized and enumeration-based algorithms to solve the problems. As problem solving techniques, the randomized algorithm makes the random choices while enumeration-based algorithm enumerates all possible candidates for solution. The performance measures of the algorithms are the route length and computation time. The experiment on the algorithms are performed based on the chain code representation derived from established previous works of Center of Excellence for Document Analysis and Recognition (CEDAR) dataset which consists of 126 upper-case letter characters. The experimental result shows that route length of both algorithms are similar but the computation time of enumeration-based algorithm is higher than randomized algorithm. This is because enumeration-based algorithm considers all branches in route walk.
A Comprehensive Study On Handwritten Character Recognition Systemiosrjce
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Handwriting Recognition Using Deep Learning and Computer VersionNaiyan Noor
This document presents a method for handwriting recognition using deep learning and computer vision. It discusses preprocessing images by removing noise and converting to grayscale. Thresholding is used to separate darker text pixels from lighter background pixels. The image is then segmented into individual lines and words. Python libraries like TensorFlow, Spyder and Jupyter Notebook are used. The goal is to build a system that can recognize text in images and display the text to users. Future work may include recognizing cursive text and additional languages.
This document discusses text detection and character recognition from images. It begins with an introduction and then discusses the aims, objectives, motivation and problem statement. It reviews relevant literature on segmentation and recognition techniques. The document then describes the methodology used, including preprocessing, segmentation using vertical projections and connected components, and recognition using pixel counting, projections, template matching, Fourier descriptors and heuristic filters. It presents results from four experiments comparing different segmentation and recognition methods. The discussion analyzes results and limitations. The conclusion finds that segmentation works best with connected components while recognition works best with template matching, Fourier descriptors and heuristic filters.
Handwritten character recognition is one of the most challenging and ongoing areas of research in the
field of pattern recognition. HCR research is matured for foreign languages like Chinese and Japanese but
the problem is much more complex for Indian languages. The problem becomes even more complicated for
South Indian languages due to its large character set and the presence of vowels modifiers and compound
characters. This paper provides an overview of important contributions and advances in offline as well as
online handwritten character recognition of Malayalam scripts.
Artificial Neural Network For Recognition Of Handwritten Devanagari CharacterIOSR Journals
1) The document discusses recognizing handwritten Devanagari characters using artificial neural networks and zone-based feature extraction.
2) It proposes extracting features from images by dividing them into zones and calculating average pixel distances to the image and zone centroids.
3) This zone-based feature vector is then input to a feedforward neural network for character recognition.
Handwritten Character Recognition: A Comprehensive Review on Geometrical Anal...iosrjce
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Continuous speech segmentation using local adaptive thresholding technique in...TELKOMNIKA JOURNAL
Continuous speech is a form of natural human speech that is continuous without a clear boundary between words. In continuous speech recognition, a segmentation process is needed to cut the sentence at the boundary of each word. Segmentation becomes an important step because a speech can be recognized from the word segments produced by this process. The segmentation process in this study was carried out using local adaptive thresholding technique in the blocking block area method. This study aims to conduct performance comparisons for five local adaptive thresholding methods (Niblack, Sauvola, Bradley, Guanglei Xiong and Bernsen) in continuous speech segmentation to obtain the best method and optimum parameter values. Based on the results of the study, Niblack method is concluded as the best method for continuous speech segmentation in Indonesian language with the accuracy value of 95%, and the optimum parameter values for such method are window = 75 and k = 0.2.
International Journal of Research in Engineering and Science is an open access peer-reviewed international forum for scientists involved in research to publish quality and refereed papers. Papers reporting original research or experimentally proved review work are welcome. Papers for publication are selected through peer review to ensure originality, relevance, and readability.
Conversion of braille to text in English, hindi and tamil languagesIJCSEA Journal
This document describes a method for converting scanned Braille documents to text in English, Hindi, and Tamil. The Braille characters are extracted from preprocessed images using segmentation and thresholding. The dot patterns in each Braille cell are converted to number sequences and mapped to letters in the appropriate language based on standard Braille codes. The converted text can then be synthesized to speech. A keyboard interface is also proposed that allows typing Braille characters via number keys corresponding to dot positions.
An Application of Eight Connectivity based Two-pass Connected-Component Label...CSCJournals
The intrinsic noise present in the image during the acquisition phase marks the recognition of Braille dots a challenging task in Optical Braille Recognition (OBR). Further, while the Braille document is being embossed on either side in the case of Inter-Point Braille, this problem of Braille dot recognition is aggravated and it makes the differentiation between recto (convex) dots and verso (concave) dots more complex. Also, the recognition of Braille dots should be carried out by reading information recorded on both sides of paper by scanning only one side. This work proposes a novelty to circumvent this issue for distinguishing convex points from concave points even if they are adjacent to each other by using only the shadow patterns of the dots and by employing the connected component labelling using two-pass algorithm and the eight connectivity property of a pixel. Enthused by the fact that, during the acquisition phase, the reflection of light through the verso dots results in a high pixel count for them when compared to the recto dots, this technique works perfectly well with good quality Braille. Furthermore, due to the natural problems like ageing and frequent usage of the document the Braille dots tend to deteriorate resulting in the down fall of the performance of the algorithm for the Braille image. Besides to this for the recognition of the Braille cell in a Braille document with some special cases an adaptive grid construction technique has also been proposed. The results extracted reveal that the enactment of the proposed technique is much consistent and dependable and that the accuracy is very much comparable to the modern state of the art techniques.
An Improved Approach for Word Ambiguity RemovalWaqas Tariq
Word ambiguity removal is a task of removing ambiguity from a word, i.e. correct sense of word is identified from ambiguous sentences. This paper describes a model that uses Part of Speech tagger and three categories for word sense disambiguation (WSD). Human Computer Interaction is very needful to improve interactions between users and computers. For this, the Supervised and Unsupervised methods are combined. The WSD algorithm is used to find the efficient and accurate sense of a word based on domain information. The accuracy of this work is evaluated with the aim of finding best suitable domain of word. Keywords: Human Computer Interaction, Supervised Training, Unsupervised Learning, Word Ambiguity, Word sense disambiguation
BrailleOCR: An Open Source Document to Braille Converter Applicationpijush15
This presentation is actually about an Open Source application, BrailleOCR that helps to convert scanned documents to Braille and thus helps the Visually Impaired.
What is the use of this application in real life? Well, BrailleOCR is currently the only app that integrated Optical character recognition and Braille Translation together. This app will eventually help converting a lot of important documents to Braille. The project site for this project is given here
IJCA Paper: http://www.ijcaonline.org/archives/volume68/number16/11664-7254
Project site: https://code.google.com/p/brailleocr/
The app uses a four step process. Initially, we have a scanned image, which is a RGB image. The first step or the Pre-Processing step deals with conversion of a RGB image to grayscale. The 2nd step deals with Character Recognition using the Tesseract Engine. Now, the recognition step may have errors and we require post processing to correct them. The 3rd step is thus the Post-Processing step and it actually corrects errors in the previous step. The final and the most important step is the Braille Conversion step.
Parameters Optimization for Improving ASR Performance in Adverse Real World N...Waqas Tariq
From the existing research it has been observed that many techniques and methodologies are available for performing every step of Automatic Speech Recognition (ASR) system, but the performance (Minimization of Word Error Recognition-WER and Maximization of Word Accuracy Rate- WAR) of the methodology is not dependent on the only technique applied in that method. The research work indicates that, performance mainly depends on the category of the noise, the level of the noise and the variable size of the window, frame, frame overlap etc is considered in the existing methods. The main aim of the work presented in this paper is to use variable size of parameters like window size, frame size and frame overlap percentage to observe the performance of algorithms for various categories of noise with different levels and also train the system for all size of parameters and category of real world noisy environment to improve the performance of the speech recognition system. This paper presents the results of Signal-to-Noise Ratio (SNR) and Accuracy test by applying variable size of parameters. It is observed that, it is really very hard to evaluate test results and decide parameter size for ASR performance improvement for its resultant optimization. Hence, this study further suggests the feasible and optimum parameter size using Fuzzy Inference System (FIS) for enhancing resultant accuracy in adverse real world noisy environmental conditions. This work will be helpful to give discriminative training of ubiquitous ASR system for better Human Computer Interaction (HCI). Keywords: ASR Performance, ASR Parameters Optimization, Multi-Environmental Training, Fuzzy Inference System for ASR, ubiquitous ASR system, Human Computer Interaction (HCI)
A bidirectional text transcription of braille for odia, hindi, telugu and eng...eSAT Journals
This document describes a bidirectional text transcription system for converting Braille documents in Odia, Hindi, Telugu, and English to text and vice versa using image processing techniques on an FPGA. It discusses prior work on Braille recognition and text-to-speech systems. The proposed algorithm segments Braille cells from image inputs and uses a modified database to map dot patterns to letters/words in each language based on number of dots. Letter patterns are rearranged for Hindi, Telugu, and English. The database is tested and dumped onto an FPGA for hardware implementation and bidirectional conversion between Braille and text in multiple languages.
IRJET - A Survey on Recognition of Strike-Out Texts in Handwritten DocumentsIRJET Journal
This document summarizes research on recognizing strike-out or crossed-out text in handwritten documents. It discusses 4 papers that studied this problem in different languages like English, Bengali, and Devanagari. The first paper used an LSTM model to recognize struck-out English text with 11% character error rate. The second used HMMs to recognize French text crossed out with lines or waves, achieving 46-92% accuracy. The third used graph algorithms and SVM to detect and remove Bengali strike-outs, obtaining 95% accuracy. The fourth used decision trees to classify English forensic documents, identifying 48% of crossed texts correctly. Overall the document reviews approaches to strike-out text recognition and removal.
Pattern Recognition of Japanese Alphabet Katakana Using Airy Zeta FunctionEditor IJCATR
Character recognition is one of common pattern recognition study. There are many object used in pattern recognition, such
as Japanese alphabet character, which is a very complex character compared to common Roman character. This research focus on
pattern recognition of Japanese character handwriting, Katakana. The pattern recognition process of a letter of the alphabet uses Airy
Zeta Function, with its input file is a .bmp file. User can write directly on an input device of the system. The testing of the system
examines 460 letter characters. The first testing that examines 230 characters result in an accuracy of 55,65%, whilst the second testing
that examines 460 characters produces an accuracy of 64,56% in recognizing the letters. These accuracy are much determined by the
quantity of training. The approach of pattern recognition is a statistical approach, where more pattern of letters are trained and saved as
a reference, more intelligent the system . The implementation of Airy zeta function methods in recognizing Japanese letter pattern is
able to produce high accuracy level.
The presentation will describe an algorithm through which one can recognize Devanagari Characters. Devanagari is the script in which Hindi is represented. This algorithm
could automatically segment character from the image of Devenagari text and then recognize them.
For extracting the individual characters from the image of Devanagari text, algorithm segmented the image several
times using the vertical and horizontal projection.
The algorithm starts with first segmenting the lines separately from the document by taking horizontal projection and then the line
into words by taking vertical projection of the line. Another step which is particular to the separation of
Devanagari characters was required and was done by first removing the header line by finding horizontal projection
of each word. The characters can then be extracted by vertical projection of the word without the header line.
Algorithm uses a Kohonen Neural Netowrk for the recognition task. After the separation of the characters from the
image, the image matrix was then downsampled to bring it down to a fixed size so as to make the recognition
size independent. The matrix can then be fed as input neurons to the Kohonen Neural Network and the winning neuron is
found which identifies the recognized the character. This information in Kohonen Neural Network was stored
earlier during the training phase of the neural network. For this, we first assigned random weights from input neurons
to output neurons and then for each training set, the winning neuron was calculated by finding the maximum
output produced by the neurons. The wights for this winning neuron were then adjusted so that it responds to this
pattern more strongly the next time.
The document discusses optical character recognition for Urdu handwriting. It introduces OCR and its applications. It then discusses earlier work on OCR systems that were font-specific. The document outlines the steps in OCR including image acquisition, preprocessing, segmentation, feature extraction, classification, and recognition. It provides an overview of the Urdu script and its variations. The document then summarizes research conducted on recognizing offline isolated Urdu characters using moment invariants and support vector machines. Other works discussed include an online and offline OCR system for Urdu using a segmentation-free approach, classifying Urdu ligatures using convolutional neural networks, and a segmentation-based approach for Urdu Nastaliq script recognition.
DEVNAGARI DOCUMENT SEGMENTATION USING HISTOGRAM APPROACHijcseit
This document summarizes a research paper on Devnagari document segmentation using a histogram approach. It discusses challenges in segmenting the Devnagari script used for several Indian languages. A simple algorithm is proposed using horizontal and vertical histograms to segment documents into lines, words and characters. The algorithm achieves near 100% accuracy for line segmentation but lower accuracy for word and character segmentation due to complexities in the Devnagari script. Future work is needed to improve character segmentation handling connected and modified characters.
This paper presents a new multi-tier holistic approach for recognizing Urdu text written in Nastaliq script. It first identifies special ligatures like dots, tay, hamza and mad from base ligatures. It then associates the special ligatures with neighboring base ligatures. Features are extracted from the ligatures and special ligature-base ligature associations. These features are input to a neural network that recognizes the ligatures in three steps: 1) identifying special ligatures, 2) associating them with base ligatures, and 3) recognizing the base ligatures. The system was tested on 200 ligatures with 100% accuracy for ligatures in its training set and closest match classification for new ligatures.
The Heuristic Extraction Algorithms for Freeman Chain Code of Handwritten Cha...Waqas Tariq
Handwriting character recognition (HCR) is the ability of a computer to receive and interpret handwritten input. In HCR, there are many representation schemes and one of them is Freeman chain code (FCC). Chain code is a sequence of code direction of a characters and connection to a starting point which is often used in image processing. The main problem in representing character using FCC that it is depends on the starting points. Unfortunately, the study about FCC extraction using one continuous route and to minimizing the length of chain code to FCC from a thinned binary image (TBI) have not been widely explored. To solve this problem, heuristic algorithms are proposed to extract the FCC that is correctly representing the characters. This paper proposes two heuristics algorithm that are based on randomized and enumeration-based algorithms to solve the problems. As problem solving techniques, the randomized algorithm makes the random choices while enumeration-based algorithm enumerates all possible candidates for solution. The performance measures of the algorithms are the route length and computation time. The experiment on the algorithms are performed based on the chain code representation derived from established previous works of Center of Excellence for Document Analysis and Recognition (CEDAR) dataset which consists of 126 upper-case letter characters. The experimental result shows that route length of both algorithms are similar but the computation time of enumeration-based algorithm is higher than randomized algorithm. This is because enumeration-based algorithm considers all branches in route walk.
A Comprehensive Study On Handwritten Character Recognition Systemiosrjce
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Handwriting Recognition Using Deep Learning and Computer VersionNaiyan Noor
This document presents a method for handwriting recognition using deep learning and computer vision. It discusses preprocessing images by removing noise and converting to grayscale. Thresholding is used to separate darker text pixels from lighter background pixels. The image is then segmented into individual lines and words. Python libraries like TensorFlow, Spyder and Jupyter Notebook are used. The goal is to build a system that can recognize text in images and display the text to users. Future work may include recognizing cursive text and additional languages.
This document discusses text detection and character recognition from images. It begins with an introduction and then discusses the aims, objectives, motivation and problem statement. It reviews relevant literature on segmentation and recognition techniques. The document then describes the methodology used, including preprocessing, segmentation using vertical projections and connected components, and recognition using pixel counting, projections, template matching, Fourier descriptors and heuristic filters. It presents results from four experiments comparing different segmentation and recognition methods. The discussion analyzes results and limitations. The conclusion finds that segmentation works best with connected components while recognition works best with template matching, Fourier descriptors and heuristic filters.
Handwritten character recognition is one of the most challenging and ongoing areas of research in the
field of pattern recognition. HCR research is matured for foreign languages like Chinese and Japanese but
the problem is much more complex for Indian languages. The problem becomes even more complicated for
South Indian languages due to its large character set and the presence of vowels modifiers and compound
characters. This paper provides an overview of important contributions and advances in offline as well as
online handwritten character recognition of Malayalam scripts.
Artificial Neural Network For Recognition Of Handwritten Devanagari CharacterIOSR Journals
1) The document discusses recognizing handwritten Devanagari characters using artificial neural networks and zone-based feature extraction.
2) It proposes extracting features from images by dividing them into zones and calculating average pixel distances to the image and zone centroids.
3) This zone-based feature vector is then input to a feedforward neural network for character recognition.
Handwritten Character Recognition: A Comprehensive Review on Geometrical Anal...iosrjce
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Continuous speech segmentation using local adaptive thresholding technique in...TELKOMNIKA JOURNAL
Continuous speech is a form of natural human speech that is continuous without a clear boundary between words. In continuous speech recognition, a segmentation process is needed to cut the sentence at the boundary of each word. Segmentation becomes an important step because a speech can be recognized from the word segments produced by this process. The segmentation process in this study was carried out using local adaptive thresholding technique in the blocking block area method. This study aims to conduct performance comparisons for five local adaptive thresholding methods (Niblack, Sauvola, Bradley, Guanglei Xiong and Bernsen) in continuous speech segmentation to obtain the best method and optimum parameter values. Based on the results of the study, Niblack method is concluded as the best method for continuous speech segmentation in Indonesian language with the accuracy value of 95%, and the optimum parameter values for such method are window = 75 and k = 0.2.
International Journal of Research in Engineering and Science is an open access peer-reviewed international forum for scientists involved in research to publish quality and refereed papers. Papers reporting original research or experimentally proved review work are welcome. Papers for publication are selected through peer review to ensure originality, relevance, and readability.
Conversion of braille to text in English, hindi and tamil languagesIJCSEA Journal
This document describes a method for converting scanned Braille documents to text in English, Hindi, and Tamil. The Braille characters are extracted from preprocessed images using segmentation and thresholding. The dot patterns in each Braille cell are converted to number sequences and mapped to letters in the appropriate language based on standard Braille codes. The converted text can then be synthesized to speech. A keyboard interface is also proposed that allows typing Braille characters via number keys corresponding to dot positions.
An Application of Eight Connectivity based Two-pass Connected-Component Label...CSCJournals
The intrinsic noise present in the image during the acquisition phase marks the recognition of Braille dots a challenging task in Optical Braille Recognition (OBR). Further, while the Braille document is being embossed on either side in the case of Inter-Point Braille, this problem of Braille dot recognition is aggravated and it makes the differentiation between recto (convex) dots and verso (concave) dots more complex. Also, the recognition of Braille dots should be carried out by reading information recorded on both sides of paper by scanning only one side. This work proposes a novelty to circumvent this issue for distinguishing convex points from concave points even if they are adjacent to each other by using only the shadow patterns of the dots and by employing the connected component labelling using two-pass algorithm and the eight connectivity property of a pixel. Enthused by the fact that, during the acquisition phase, the reflection of light through the verso dots results in a high pixel count for them when compared to the recto dots, this technique works perfectly well with good quality Braille. Furthermore, due to the natural problems like ageing and frequent usage of the document the Braille dots tend to deteriorate resulting in the down fall of the performance of the algorithm for the Braille image. Besides to this for the recognition of the Braille cell in a Braille document with some special cases an adaptive grid construction technique has also been proposed. The results extracted reveal that the enactment of the proposed technique is much consistent and dependable and that the accuracy is very much comparable to the modern state of the art techniques.
Static-gesture word recognition in Bangla sign language using convolutional n...TELKOMNIKA JOURNAL
Sign language is the communication process of people with hearing impairments. For hearing-impaired communication in Bangladesh and parts of India, Bangla sign language (BSL) is the standard. While Bangla is one of the most widely spoken languages in the world, there is a scarcity of research in the field of BSL recognition. The few research works done so far focused on detecting BSL alphabets. To the best of our knowledge, no work on detecting BSL words has been conducted till now for the unavailability of BSL word dataset. In this research, a small static-gesture word dataset has been developed, and a deep learning-based method has been introduced that can detect BSL static-gesture words from images. The dataset, “BSLword” contains 30 static-gesture BSL words with 1200 images for training.
The training is done using a multi-layered convolutional neural network with the Adam optimizer. OpenCV is used for image processing and TensorFlow is used to build the deep learning models. This system can recognize BSL static-gesture words with 92.50% accuracy on the word dataset.
OFF-LINE ARABIC HANDWRITTEN WORDS SEGMENTATION USING MORPHOLOGICAL OPERATORSsipij
The main aim of this study is the assessment and discussion of a model for hand-written Arabic through
segmentation. The framework is proposed based on three steps: pre-processing, segmentation, and
evaluation. In the pre-processing step, morphological operators are applied for Connecting Gaps (CGs) in
written words. Gaps happen when pen lifting-off during writing, scanning documents, or while converting
images to binary type. In the segmentation step, first removed the small diacritics then bounded a
connected component to segment offline words. Huge data was utilized in the proposed model for applying
a variety of handwriting styles so that to be more compatible with real-life applications. Consequently, on
the automatic evaluation stage, selected randomly 1,131 images from the IESK-ArDB database, and then
segmented into sub-words. After small gaps been connected, the model performance evaluation had been
reached 88% against the standard ground truth of the database. The proposed model achieved the highest
accuracy when compared with the related works.
Off-Line Arabic Handwritten Words Segmentation using Morphological Operatorssipij
The document presents a model for segmenting offline Arabic handwritten words using morphological operators. It has three main stages: pre-processing, segmentation, and evaluation. In pre-processing, morphological operators like dilation and erosion are used to connect gaps in words caused by pen lifting or scanning. Segmentation uses connected component analysis to bound words into sub-words. The model is tested on over 1,100 images from the IESK-ArDB database, achieving 88% accuracy in segmenting words into sub-words by connecting small gaps. This compares favorably to previous related works on Arabic handwriting segmentation.
Gesture Acquisition and Recognition of Sign LanguageIRJET Journal
The document discusses sign language recognition techniques. It begins with an introduction to sign languages and issues faced by deaf communities in communication. It then reviews recent work in sign language recognition, covering approaches used like hand tracking, feature extraction, and classification methods. Finally, it discusses existing challenges and future research opportunities in sign language recognition.
GEOMETRIC CORRECTION FOR BRAILLE DOCUMENT IMAGEScscpconf
Braille system has been used by the visually impaired people for reading.The shortage of Braille books has caused a need for conversion of Braille to text. This paper addresses the geometric correction of a Braille document images. Due to the standard measurement of the Braille cells, identification of Braille characters could be achieved by simple cell overlapping procedure. The standard measurement varies in a scaled document and fitting of the cells become difficult if the
document is tilted. This paper proposes a line fitting algorithm for identifying the tilt (skew) angle. The horizontal and vertical scale factor is identified based on the ratio of distance between characters to the distance between dots. These are used in geometric transformation matrix for correction. Rotation correction is done prior to scale correction. This process aids in increased accuracy. The results for various Braille documents are tabulated.
GEOMETRIC CORRECTION FOR BRAILLE DOCUMENT IMAGEScsandit
Braille system has been used by the visually impaired people for reading.The shortage of Braille
books has caused a need for conversion of Braille to text. This paper addresses the geometric
correction of a Braille document images. Due to the standard measurement of the Braille cells,
identification of Braille characters could be achieved by simple cell overlapping procedure. The
standard measurement varies in a scaled document and fitting of the cells become difficult if the
document is tilted. This paper proposes a line fitting algorithm for identifying the tilt (skew)
angle. The horizontal and vertical scale factor is identified based on the ratio of distance
between characters to the distance between dots. These are used in geometric transformation
matrix for correction. Rotation correction is done prior to scale correction. This process aids in
increased accuracy. The results for various Braille documents are tabulated.
This paper presents an approach to recognize off-line Bangla numeral. Today there are many OCR used to recognize
Bangla numeral. The recognition of handwritten character is still a challenging work in the field of pattern recognition. Numeral
recognition in pattern recognition is the process to identify the given character according to the predefined character set. The difficulties
of recognition of handwritten Bangla numeral are that they are different in shapes and sizes which are much curved in nature. . We try
to establish a process to recognize such handwritten Bangla numerals having different shape and size. The input scanned image is first
to be binarized. Then we have segmented all the ten digits of Bangla numerals to identify each and individual digit from a scanned
image. We have used line segmentation to extract the feature from each numeral based on templates. A high correlation coefficient
method provides a successful match between the test data and training data.
SEGMENTATION OF CHARACTERS WITHOUT MODIFIERS FROM A PRINTED BANGLA TEXTcscpconf
This document discusses the segmentation of printed Bangla characters without modifiers for optical character recognition systems. It begins with an introduction to OCR systems and Bangla script. The main steps of an OCR system are then outlined, with a focus on the segmentation step. Line, word and character segmentation algorithms are described in detail along with figures to illustrate the steps. The goal is to properly segment individual characters for recognition.
SENTIMENT ANALYSIS IN MYANMAR LANGUAGE USING CONVOLUTIONAL LSTM NEURAL NETWORKijnlc
In recent years, there has been an increasing use of social media among people in Myanmar and writing review on social media pages about the product, movie, and trip are also popular among people. Moreover, most of the people are going to find the review pages about the product they want to buy before deciding whether they should buy it or not. Extracting and receiving useful reviews over interesting products is very important and time consuming for people. Sentiment analysis is one of the important processes for extracting useful reviews of the products. In this paper, the Convolutional LSTM neural network architecture is proposed to analyse the sentiment classification of cosmetic reviews written in Myanmar Language. The paper also intends to build the cosmetic reviews dataset for deep learning and sentiment lexicon in Myanmar Language.
Sentiment Analysis In Myanmar Language Using Convolutional Lstm Neural Networkkevig
In recent years, there has been an increasing use of social media among people in Myanmar and writing
review on social media pages about the product, movie, and trip are also popular among people. Moreover,
most of the people are going to find the review pages about the product they want to buy before deciding
whether they should buy it or not. Extracting and receiving useful reviews over interesting products is very
important and time consuming for people. Sentiment analysis is one of the important processes for extracting
useful reviews of the products. In this paper, the Convolutional LSTM neural network architecture is
proposed to analyse the sentiment classification of cosmetic reviews written in Myanmar Language. The
paper also intends to build the cosmetic reviews dataset for deep learning and sentiment lexicon in Myanmar
Language.
This document discusses the development of an Indian Sign Language recognition system called SignReco. It begins with an abstract describing the challenges faced by deaf individuals communicating with others without translation and the benefits of a system that can recognize sign language. The paper then provides background on sign language and the goals of the proposed system, which is to classify and recognize Indian Sign Language in real-time using CNN and neural networks. A literature review covers prior work on sign language recognition systems. The proposed system's workflow and modules for model creation, language translation and app development are described. It concludes that the survey helped in developing an effective approach for an Indian Sign Language recognition system using CNN to improve accuracy.
This document is a project report submitted by Mohammad Saiful Islam for a CMPUT 551 course on December 21st, 2010 regarding Bengali handwritten digit recognition using support vector machines. The report discusses building a dataset of Bengali digits written by the author, preprocessing and feature extraction steps, and using a multiclass support vector machine with different kernels for classification. The author hypothesizes that SVM will perform well, RBF kernels will improve performance over linear and polynomial kernels, and using raw pixel values can achieve good accuracy, though testing on different writers may reduce performance. Experiments are planned to test these hypotheses using the collected dataset.
The document describes a project to develop a real-time sign language detection system using computer vision and deep learning techniques. The researchers collected over 500 images of 5 different signs and trained a convolutional neural network model using transfer learning with a pre-trained SSD MobileNet V2 model. The model takes input from a webcam video stream and classifies each frame in real-time to detect the sign language. Some key applications of this system include improving communication for deaf individuals and teaching sign language. The researchers achieved reliable detection results under controlled lighting conditions and aim to expand the dataset and model capabilities in future work.
An effective approach to offline arabic handwriting recognitionijaia
Segmentation is the most challenging part of the Arabic handwriting recognition, due to the unique
characteristics of Arabic writing that allows the same shape to denote different characters. In this paper,
an off-line Arabic handwriting recognition system is proposed. The processing details are presented in
three main stages. Firstly, the image is skeletonized to one pixel thin. Secondly, transfer each diagonally
connected foreground pixel to the closest horizontal or vertical line. Finally, these orthogonal lines are
coded as vectors of unique integer numbers; each vector represents one letter of the word. In order to
evaluate the proposed techniques, the system has been tested on the IFN/ENIT database, and the
experimental results show that our method is superior to those methods currently available.
SIGN LANGUAGE RECOGNITION USING MACHINE LEARNINGIRJET Journal
1. The document describes a study on developing a real-time sign language recognition system using machine learning. The system captures hand gestures using a webcam and identifies the region of interest to predict the sign.
2. Convolutional neural networks are used to train the model to classify signs. Related works that also use CNNs and other machine learning techniques for sign language recognition from images are discussed.
3. The proposed system aims to make communication easier for deaf and mute people by automatically translating signs to text in real-time without requiring an expert translator.
Online Hand Written Character RecognitionIOSR Journals
This document discusses online handwritten character recognition. It begins by describing the differences between online and offline recognition systems. Online systems capture stroke order and timing information while writing, while offline systems analyze static images. The document then discusses challenges in recognition like variability between writers. It presents several previous works in online handwriting recognition. The document proposes a method for online recognition that uses shape, pixel density, and stroke movement template matching to identify characters. It describes preprocessing input and generating training templates to match against. Overall, the document outlines challenges in online handwriting recognition and proposes a template matching approach to address these challenges.
Live Sign Language Translation: A SurveyIRJET Journal
The document discusses various approaches that have been used for live sign language translation. It reviews 20 research papers that used techniques like convolutional neural networks, support vector machines, k-nearest neighbors, and LSTM networks to classify hand gestures and translate sign language into text with varying levels of accuracy between 62.3% to 99.9%. Deep learning models using CNNs and LSTMs achieved the highest accuracy compared to traditional classifiers. The paper aims to help other researchers in the field understand past approaches and how to potentially improve sign language translation systems.
Device for text to speech production and to braille scriptIAEME Publication
The document describes a proposed system to convert text to both speech and Braille script for blind or deaf individuals. The system would take an image of text as input, perform image processing techniques like enhancement, filtering, and edge detection, then segment and recognize characters. The recognized text would be converted to speech output using text-to-speech synthesis or to Braille script by mapping characters to Braille codes and outputting to a tactile display. The goal is to make learning materials more accessible for blind or deaf individuals by converting textbook images to audio or Braille formats.
Similar to OPTICAL BRAILLE TRANSLATOR FOR SINHALA BRAILLE SYSTEM: PAPER COMMUNICATION TOOL BETWEEN VISION IMPAIRED AND SIGHTED PERSONS (20)
ACEP Magazine edition 4th launched on 05.06.2024Rahul
This document provides information about the third edition of the magazine "Sthapatya" published by the Association of Civil Engineers (Practicing) Aurangabad. It includes messages from current and past presidents of ACEP, memories and photos from past ACEP events, information on life time achievement awards given by ACEP, and a technical article on concrete maintenance, repairs and strengthening. The document highlights activities of ACEP and provides a technical educational article for members.
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...IJECEIAES
Climate change's impact on the planet forced the United Nations and governments to promote green energies and electric transportation. The deployments of photovoltaic (PV) and electric vehicle (EV) systems gained stronger momentum due to their numerous advantages over fossil fuel types. The advantages go beyond sustainability to reach financial support and stability. The work in this paper introduces the hybrid system between PV and EV to support industrial and commercial plants. This paper covers the theoretical framework of the proposed hybrid system including the required equation to complete the cost analysis when PV and EV are present. In addition, the proposed design diagram which sets the priorities and requirements of the system is presented. The proposed approach allows setup to advance their power stability, especially during power outages. The presented information supports researchers and plant owners to complete the necessary analysis while promoting the deployment of clean energy. The result of a case study that represents a dairy milk farmer supports the theoretical works and highlights its advanced benefits to existing plants. The short return on investment of the proposed approach supports the paper's novelty approach for the sustainable electrical system. In addition, the proposed system allows for an isolated power setup without the need for a transmission line which enhances the safety of the electrical network
Optimizing Gradle Builds - Gradle DPE Tour Berlin 2024Sinan KOZAK
Sinan from the Delivery Hero mobile infrastructure engineering team shares a deep dive into performance acceleration with Gradle build cache optimizations. Sinan shares their journey into solving complex build-cache problems that affect Gradle builds. By understanding the challenges and solutions found in our journey, we aim to demonstrate the possibilities for faster builds. The case study reveals how overlapping outputs and cache misconfigurations led to significant increases in build times, especially as the project scaled up with numerous modules using Paparazzi tests. The journey from diagnosing to defeating cache issues offers invaluable lessons on maintaining cache integrity without sacrificing functionality.
Presentation of IEEE Slovenia CIS (Computational Intelligence Society) Chapte...University of Maribor
Slides from talk presenting:
Aleš Zamuda: Presentation of IEEE Slovenia CIS (Computational Intelligence Society) Chapter and Networking.
Presentation at IcETRAN 2024 session:
"Inter-Society Networking Panel GRSS/MTT-S/CIS
Panel Session: Promoting Connection and Cooperation"
IEEE Slovenia GRSS
IEEE Serbia and Montenegro MTT-S
IEEE Slovenia CIS
11TH INTERNATIONAL CONFERENCE ON ELECTRICAL, ELECTRONIC AND COMPUTING ENGINEERING
3-6 June 2024, Niš, Serbia
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECTjpsjournal1
The rivalry between prominent international actors for dominance over Central Asia's hydrocarbon
reserves and the ancient silk trade route, along with China's diplomatic endeavours in the area, has been
referred to as the "New Great Game." This research centres on the power struggle, considering
geopolitical, geostrategic, and geoeconomic variables. Topics including trade, political hegemony, oil
politics, and conventional and nontraditional security are all explored and explained by the researcher.
Using Mackinder's Heartland, Spykman Rimland, and Hegemonic Stability theories, examines China's role
in Central Asia. This study adheres to the empirical epistemological method and has taken care of
objectivity. This study analyze primary and secondary research documents critically to elaborate role of
china’s geo economic outreach in central Asian countries and its future prospect. China is thriving in trade,
pipeline politics, and winning states, according to this study, thanks to important instruments like the
Shanghai Cooperation Organisation and the Belt and Road Economic Initiative. According to this study,
China is seeing significant success in commerce, pipeline politics, and gaining influence on other
governments. This success may be attributed to the effective utilisation of key tools such as the Shanghai
Cooperation Organisation and the Belt and Road Economic Initiative.
A SYSTEMATIC RISK ASSESSMENT APPROACH FOR SECURING THE SMART IRRIGATION SYSTEMSIJNSA Journal
The smart irrigation system represents an innovative approach to optimize water usage in agricultural and landscaping practices. The integration of cutting-edge technologies, including sensors, actuators, and data analysis, empowers this system to provide accurate monitoring and control of irrigation processes by leveraging real-time environmental conditions. The main objective of a smart irrigation system is to optimize water efficiency, minimize expenses, and foster the adoption of sustainable water management methods. This paper conducts a systematic risk assessment by exploring the key components/assets and their functionalities in the smart irrigation system. The crucial role of sensors in gathering data on soil moisture, weather patterns, and plant well-being is emphasized in this system. These sensors enable intelligent decision-making in irrigation scheduling and water distribution, leading to enhanced water efficiency and sustainable water management practices. Actuators enable automated control of irrigation devices, ensuring precise and targeted water delivery to plants. Additionally, the paper addresses the potential threat and vulnerabilities associated with smart irrigation systems. It discusses limitations of the system, such as power constraints and computational capabilities, and calculates the potential security risks. The paper suggests possible risk treatment methods for effective secure system operation. In conclusion, the paper emphasizes the significant benefits of implementing smart irrigation systems, including improved water conservation, increased crop yield, and reduced environmental impact. Additionally, based on the security analysis conducted, the paper recommends the implementation of countermeasures and security approaches to address vulnerabilities and ensure the integrity and reliability of the system. By incorporating these measures, smart irrigation technology can revolutionize water management practices in agriculture, promoting sustainability, resource efficiency, and safeguarding against potential security threats.
Understanding Inductive Bias in Machine LearningSUTEJAS
This presentation explores the concept of inductive bias in machine learning. It explains how algorithms come with built-in assumptions and preferences that guide the learning process. You'll learn about the different types of inductive bias and how they can impact the performance and generalizability of machine learning models.
The presentation also covers the positive and negative aspects of inductive bias, along with strategies for mitigating potential drawbacks. We'll explore examples of how bias manifests in algorithms like neural networks and decision trees.
By understanding inductive bias, you can gain valuable insights into how machine learning models work and make informed decisions when building and deploying them.
Using recycled concrete aggregates (RCA) for pavements is crucial to achieving sustainability. Implementing RCA for new pavement can minimize carbon footprint, conserve natural resources, reduce harmful emissions, and lower life cycle costs. Compared to natural aggregate (NA), RCA pavement has fewer comprehensive studies and sustainability assessments.
Batteries -Introduction – Types of Batteries – discharging and charging of battery - characteristics of battery –battery rating- various tests on battery- – Primary battery: silver button cell- Secondary battery :Ni-Cd battery-modern battery: lithium ion battery-maintenance of batteries-choices of batteries for electric vehicle applications.
Fuel Cells: Introduction- importance and classification of fuel cells - description, principle, components, applications of fuel cells: H2-O2 fuel cell, alkaline fuel cell, molten carbonate fuel cell and direct methanol fuel cells.
Using recycled concrete aggregates (RCA) for pavements is crucial to achieving sustainability. Implementing RCA for new pavement can minimize carbon footprint, conserve natural resources, reduce harmful emissions, and lower life cycle costs. Compared to natural aggregate (NA), RCA pavement has fewer comprehensive studies and sustainability assessments.
Harnessing WebAssembly for Real-time Stateless Streaming PipelinesChristina Lin
Traditionally, dealing with real-time data pipelines has involved significant overhead, even for straightforward tasks like data transformation or masking. However, in this talk, we’ll venture into the dynamic realm of WebAssembly (WASM) and discover how it can revolutionize the creation of stateless streaming pipelines within a Kafka (Redpanda) broker. These pipelines are adept at managing low-latency, high-data-volume scenarios.
OPTICAL BRAILLE TRANSLATOR FOR SINHALA BRAILLE SYSTEM: PAPER COMMUNICATION TOOL BETWEEN VISION IMPAIRED AND SIGHTED PERSONS
1. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
DOI: 10.5121/ijma.2018.10303 29
OPTICAL BRAILLE TRANSLATOR FOR SINHALA BRAILLE
SYSTEM: PAPER COMMUNICATION TOOL BETWEEN
VISION IMPAIRED AND SIGHTED PERSONS
T. D. S. H. Perera, and W. K. I. L. Wanniarachchi*
Department of Physics, University of Sri Jayewardenepura, Nugegoda, Sri Lanka
ABSTRACT
In this paper we proposed a system; Optical Braille Translator (OBT), that identify Sinhala Braille
characters in single sided Braille document and translates to Sinhala language. This system also capable of
identifying Grade1 English Braille characters, numbers, capital letters and some words in Grade 2
English Braille system. Image processing techniques were used to developed the proposed system in
MATLAB environment. The translated text displayed in a word application as the final outcome.
Performance evaluation results reflect that the proposed method can recognize Braille characters and
translated to user selected language either Sinhala or English efficiently, over 99% of accuracy.
KEYWORDS
Braille, Braille Recognition, Image Processing, Optical Recognition, Sinhala
1. INTRODUCTION
According to the thirteenth Census of Population and Housing survey which was conducted on
17th July, 2001 in Sri Lanka, the number of visually impaired people has reached 69,096[1]. Out
of that 35,419 were males and 33,677 were females. Among them 10,947 people out of the
69,096 were totally blind. Therefore, it is necessary to provide support those people with
intelligent systems and technologies to improve communication and interaction with each other
and with non-blind people. The major senses used by visual impaired people are hearing and
touch feelings.The most famous communication system for blind people is the Braille system
which depends on the sense of the touch of a fingertip. Braille is a system that allows visually
impaired people to read through touch using a series of raised dots on special papers which can
only be read using fingers.Braille is not a language and these Braille characters are used to
specify character in any language[2].
Braille coding system made up of different type of characters which are also called “cells”. Each
Braille character or a “cell” is made up of six dot positions arranged as two columns of three dots
to form a rectangular shape. A dot may be raised at any of these six positions to form sixty-four
combinations including the combination which no dots are raised. Positions of these dots are
universally numbered 1 to 3 from top to bottom on the left, and 4 to 6 from top to bottom on the
right[3]. The dimension of a Braille dot, distance between dots in a cell and distance between
cells have been set according to the tactile resolution of a fingertip[3]. The horizontal and vertical
distance between dots in a cell and distance between cells in a word and inter line distance also
specified by the Library of Congress. Here, dot height is approximately 0.02 inches (0.5mm), the
horizontal and vertical spacing between dot centers within a Braille cell is approximately 0.1
inches (2.5mm), the blank space between dots on adjacent cells is approximately 0.15 inches
(3.75mm) horizontally and 0.2 inches (5.0mm) vertically. A standard Braille page is 11 inches by
11.5 inches and typically has a maximum of 34 to 40 Braille cells per line and 25 lines per
page[4]. The Braille has been adapted to write many different languages including Sinhala, also it
2. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
30
is used for musical and mathematical notation. Sinhala and English Braille letters are read from
left to write.
When complexity of Sinhala Braille is compared with the English Braille, Sinhala Braille system
can be categorizing as a grade 01 Braille system. Because Sinhala Braille having one to one
transcription with Sinhala letters. Rarely two Braille characters are used to represent single
Sinhala letter such as “ඏ, ඐ, ඍ andඎ”. When the Sinhala Braille system is compared with the
Sinhala language, some characters are missing in the Braille system. For an example, when the
word “අ මා” is written using Sinhala Braille “⠁⠍⠈⠍⠜” which look likes “අ මආ”. Because
character “◌ා” is not used in Sinhala Braille system. Therefore, pronunciation sound of the
Sinhala word is used to write that word in the Sinhala Braille system.
However, most people in the society cannot understand Braille. Paper communication between
the visually impaired people and non-blind people have become a problem that need to be
addressed. Therefore, translating Braille into Sinhala or any other languages enable
communication with people in society. In this research work, we developed a Braille translator
system which can translate Braille characters into Sinhala language. The prosed system has been
improved to identify grade1 English Braille characters, numbers, capital letters and some words
in grade 2 English Braille system. An image of a single sided Braille paper is taken as the input to
the system. Evaluation of the performance of the system showed that it can recognize Braille
characters and translating to Sinhala/English language over 99% of accuracy.The developed
system uses simple image processing techniques of low computational power and performs well
over the other published work. In the paper we comprehensively discuss the image processing
methods that we used to develop optical Braille translator (OBT).
2. PREVIOUS WORK
In the literature there are many researches have been carried out for Braille character recognition
based on image processing techniques. In the paper “Smart Braille System Recognizer”, authors
claimed that the developed system can recognize characters in single side Braille document with
94.39% accuracy. Image acquisition stage, image pre-processing, modified image segmentation,
feature extraction, and character recognition based on image processing techniques were the main
staged of their work. At the image acquisition step authors used flat-bed scanner to obtain images
of single side Braille documents [3]. On their work, J. Li et.al., optical Braille recognition system
used normal scanner to aquire the input Braille documents’ images. Geometrical corrections were
applied in preprosessing stages. Haar wavelet feature extraction and Support Vector Machine
classification techniques were perforemed on croped sub images ofBraille dots for identification.
Identified Braille cells were converted to English language with aid of searching algorithm.
Authors claims that the developed method is in acceptable level for Braille extraction [5]. M.
Wajid et. al. developed Braille to Urdu language translation system based on image processing
using MATLAB. After the pre-processingand segmentation steps,a 3×2 matrix was generated
according to the dot pattermn available in a Braille cell. Here, they used a threshold value where
the number of white pixels in selected region of the Braille cell greater than the threshold then
that element in the matrix represent a Braille dot. Generation of this pattern matrix in terms of 0’s
and 1’s was used to link corresponding letters in Urdu language [6]. L. Wonget. al. proposed a
Braille recognition system based on image processing and probabilistic neural network. The
statistics of performance evaluation of the system shows that the accuracy is 99% [7]. K.P.S.G.
Sugirthaet. al. proposed a method of translating braille code into English language.In their work,
simple flat-bet scanner used to acquire image of the braille documents. Then image pre-
processing steps including expel alignment, gray scaling, thresholding and dilation were
performed on the subjected image. Author claimed that in the segmentation step of their work
they considered the Euclidean distance to evolve new technique to recognize the characters[8].
3. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
31
The research work carried out by the E. Jacinto Gómez et. al. has clamied thet they have used
very unique method to identify braille characters. In their work they used circle hough transform
method to identify dots in image of the braille document[9]. In the paper “Braille Character
Recognition Using Associtive Memory” authorsS. H. Khaled et. al. has translated braille
document into English language and voice. Their resech consisting two main stages,
preprocessing stage and recognition stage. In the second stage Modify Multy-Connect
Architecture (MMCA) and Modify Bidirectional Associative Memory (MBAM) algorithms were
implemented.MMCA algorithm has achived araerage accuracy for correct letter is 98.26%,
average correct word was 95.11% and average processing time around 11.5 seconds per page.
MBAM algorithm achived aearage accuracy for correct letters is 91.87%, average accuracy for
correct word was 51.26% and average processing time around 3.4 seconds per page[10].There are
very few work previously done for recognizing Sinhala Braille letters.The research work carried
out translating Sinhala text document into Braille by Soma Chatterjee [11];the proposed system
can convert MS word-based Unicode Sinhala document to Braille. Recently in 2016, N.M.T De
Silva et. al. proposed a system to convert Braille to Sinhala characters. On their work, K-nearest
neighbor classification method was used for identification of Braille characters. The outcome can
implement Unicode mapping for 52 Sinhala characters and 10 numbers. The test results reflected
that the developed system could translate Braille characters to Sinhala language with 91.4%
accuracy [12].
3. METHOD
This research work is based on image processing techniques where computer algorithms
implemented on MATLAB environment. Optical Braille Translator (OBT) is a system that
identify the Braille characters in an image of a one side Braille document and translate them into
corresponding natural language. Image can be a scanned image of a Braille document or color
image taken by a camera. Finally,each and every extractedBraille character is translated into
corresponding letter in the Sinhalalanguage. Also, OBT has the ability to identify Grade 01
English Braille characters and some words (e.g. and, for, of, the, with…etc.) in English grade 2
Braille[13]–[15].Furthermore, the OBT system is capable of finding numbers in both Sinhala and
English Braille document and capital letters in grade 1 English Braille document. Final output is
written toa Microsoft Word document[16]. A simple graphical user interface has been designed
for user interaction. The main steps in the developed OBT system is shown in the figure 1.Each of
the step according to the system flow chart in figure 1 is discussed comprehensively in the
following sections.
Figure 1: Flow chart of the developed OBT system
3.1 Image Acquisition
4. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
32
Image acquisition is the manual step of this system and the accuracy and time taken to translate
the Braille document into a natural language mainly depend on the quality of the acquired image.
In this research,two different types of methods were used to get an image of a Braille document:
A hand written Braille document (a Braille document where Braille dots were written by pen) was
scanned by a normal scanner, and an image of a Braille document was generated by a computer.
Figures 2 (a), and 2(b) show the selected images for the further steps in this system.
Figure 2: (a) Scanned image of a hand-written braille document (b) Computer generated braille document
Figure 3: Sub steps of the image pre-processing
Here, the input image is a true color RGB image (24-bit image).
3.2 Pre-Processing
Pre-processing step consists of several sub stepssuch as gray image processing, binary image
generation, angle correction and image resizing functions which are performed on the input
image. The figure 3 shows the sub steps of the pre-processing step. The acquired image in the
initial step is input to the image pre-processing routing. Then, the color image is converted into
grayscale image. Image dilation, erosion, and image subtraction were performed before
5. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
33
converting gray image into a binary image[17]. In order to obtain the edges of the image, image
dilation and erosion operations were carried out. Here, morphological disk of radius 1 structural
element [0 1 0;1 1 1; 0 1 0] was used for dilation and erosion operation[18]. Finally, edges of the
foreground objects were successfully obtained by subtracting dilated image from the eroded
image. Then the global threshold was considered to obtain the binary image. For the image
enhancement, holes filling and noise reduction techniques were used. Figure 4. (a) shows the
binary image obtained according to the global threshold which consists of noises whereas figure
4. (b) shows the hole filled binary image. In order to reduce the noises from the image, MATLAB
function “bwareaopen” was used. Here, we removed objects that have fewer than 10 connected
pixels and obtained the noise removed binary image (figure 4. (c)). Before proceed to the
segmentation process, correct alignment of the Braille characters was processed. Figure 5 shows
results of the angle correction obtained from the Radon transformation [19]. In order to decrease
the computation time for further the processing, the correct alignment image is resized to lower
resolution which consists of 480 rows while keeping original image aspect ratio.
Figure 4: Image enhancement (a) binary image, (b) Hole filled binary image (c) noise reduced binary image
Figure 5: Angle correction using Radon transformation. Initial image (left) and angle corrected resized
image (right)
6. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
34
3.3Braille Character Segmentation
Summation of pixels’ value took along the rows and columns of the noise removed binary
imagewas taken into account in order to identify Braille character cells. Figure 6. (a) and figure 6.
(b) showrow sum and the column sum of an input binary image. Row summation (Figure 6. (a))
was used to identify the average vertical distance between two rows. An example is highlighted in
red circle and average vertical distance between two Braille dots in a cell (shown in arrows).
According to the row summation shown in figure 6 (a), the smaller zero count vertical gaps
reflect the separation of two Braille dots in a given cell while the larger zero count vertical gaps
reflect the separation of adjacent character rows. Similarly, in column sum (figure 6. (b)) larger
horizontal zero count gaps represent the separation between Braille character columns while
smaller zero count gaps reflect the Braille dot separation in a given cell. Accordingly, a computer
algorithm was developed to find the approximate vertical and horizontal position of characters
automatically (see figure 7). At the end of the segmentation step, there are two sets of data points,
one represents the segmentation along the vertical direction while other represents the
segmentation points along the horizontal direction.
Figure 6: Stair graph of row sum and column sum
= = {
0 ( )
1 ( ℎ )
(1)
lx = length of the image by pixels (number of columns)
yi = horizontal projection of the pixels (summation along the rows of pixels)
= = {
0 ( )
1 ( ℎ )
(2)
ly = height of the image by pixels (number of rows)
xi = vertical projection of the pixels (summation along the columns of pixels)
(i,j = vertical and horizontal coordination of the pixels)
7. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
35
3.4 Character Extraction
In the Braille character segmentation step, the upper and lower row separation positions and the
upper and lower column separation positions were obtained. In the character extraction, first rows
were separated from the noise removed binary image according to the row separation positions.
Then the Braille characters were extracted by cropping the row character images as per the values
obtained by the column separation position. The extracted characters were resized to 21×16
matrix binary images. At the end of this process, the Braille character cells were successfully
extracted from the initial RGB input image.
Figure 7: Main steps of the Braille character segmentation process
3.5. Braille Character Regeneration
In order to increase the accuracy of Braille character recognition, we regenerated the Braille cell
image according to the extracted Braille character images. Figure 8. (a) shows the extracted
21×16 matrix binary image of a Braille character. For the regeneration process, the column width
was divided into three columns which have width of 5, 6 and 5 pixels while row height divided in
to 5, 3, 5, 3 and 5 pixels. The figure 8. (b) graphically illustrates the separate region in the Braille
character for the regeneration process. Here, a single Braille dot in a cell is represented by a 5×5
matrix element. Accordingly, there are six 5×5 matrix regions (region B) as shown in the figure
8. (c). In practical cases, we have observed that the Braille dots may not centered in these 5×5
matrix regions as shown in figure 8. (b). Hence, identification of Braille characters by a computer
system may be time consuming without regenerating the Braille cell properly. The developed
algorithm for regeneration of Braille characters is shown in figure 9. Here, the number of white
Figure 8: Braille character regeneration process Figure 9: Braille character regeneration algorithm
8. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
36
pixels in 5×8 area (shown in the figure 8. (c) region A) which covers 5×5 matrix (region B), is
taken in to account when deciding the Braille dot formation. If the number of white pixel in
region A of any region B sites, then corresponding region B site converted to white color; else
converted to black color. The final output of Braille character regeneration of a tested image is
shown in figure 8 (d) which reflects the enhanced quality image for character identification.
3.6. Braille Character Recognition
In this step, binary to decimal equivalent number is incorporated for identification of each Braille
character. As shown in figure 10., value of each midpoint of the Braille dot regions in the
regenerated Braille cell were used to generate a 6-bit binary number. The binary equivalent
decimal number can be obtained from D5×25
+D4×24
+D3×23
+D2×22
+D1×21
+D0×20
arithmetic
operation. The middle point of the Braille dot location one was taken as the least significant bit
and the middle point of the Braille dot location six assigned to the most significant bit. The
developed computer program takes the value of each midpoint located at (3,3), (11,3), (19,3),
(3,14), (11,14) and (19,14) in the regenerated Braille character image to make digital
representation of the Braille character where this digital representation 6-bit binary number
identical to corresponding Braille character. The figure 11.indicates binary equivalent decimal
numbers as a 2D array for a given input image consist of 32 (4×8) Braille characters. Hence,
corresponding binary to decimal equivalent number can relate to characters in natural languages.
The Sinhala and Grade I English letters and some of the Grade II English words with binary
equivalent decimal number for corresponding Braille character are shown in the figure 12. Here,
we used this database to decode the Braille characters in the proposed system.
The binary equivalent decimal numbers corresponding to Braille characters are shown in top of
each cell in figure 12. The corresponding Sinhala and English letters (some words in Grade II
English) are shown in black color. Green color letters (strings) represents the corresponding
English keyboard characters in “AA Amali” font type which is discussed in section 3.7.
9. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
37
Figure 12: OBT Database for Decoding Braille Characters
Figure 13: Number identification process in Braille to English translation
The proposed optical Braille translator system is also capable of identifying numbers. In Sinhala
Braille system as well as in English Braille system “⠼” Braille character (binary equivalent
decimal number is 60) is used to represent numbers. In this work, the corresponding character “#”
was assigned to represent “⠼”. If a space followed by #, then it represents the character #. If a
10. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
38
character(s) followed by the #, then it represents a number. The proposed system first converts
Braille characters in to user define language either Sinhala (see section 3.7 for more detail) or
English. Then the developed algorithm searches for “#” character by scanning identified
characters. When character “#” found, then searches for “ ” (space) in the array and all the
characters in between # and “ ” converts to the relevant number. Then the identified numbers
replaced to the correct position in the original character array. This process repeats until the end
of array is reached. In the figure 14, steps for the number identification is visualized. In grade I
English Braille uses “⢀” Braille character (binary equivalent decimal number is 32) to represent
capital letters. This Braille character is identified as “@” sign in our proposed system. Any
character followed by “@” converted to capital letters and the capital letter searching function is
developed similar to the number identification process shown in figure 13. This subroutine is
applied for Braille to English translation only.
3.7. Braille to Sinhala Translation
The OBT developed in this work capable of translating Sinhala Braille document to Sinhala
letters in “AA Amali” font [20] in a Microsoft word document. After obtaining the binary
equivalent decimal number array, corresponds to the input Braille document, identification of
natural language character is proceeded. For the identification process, the matching character to
the decimal number is selected as per the database shown in figure 12.But when converting to
Sinhala language the relevant English keyboard character or string in “AA Amali” font type is
used in MATLAB environment according to the database based on figure 12.As an example, if
the obtained binary equivalent decimal number array is [28, 25, 23, 61], then the system stores the
corresponding text array as [wd, o, r, h] in MATLAB environment. Here, we used English
Figure 14: Sinhala translated “text_arry” in MATLAB Environment
keyboard letter/string representation of “AA Amali” font type when translating decimal number
to Sinhala letter in MATLAB environment.Then the text array sent to word application where it
displays the text array as “wdorh” in selected font type “AA Amali”. Before sending the text
array into word application, number identification subroutine is called to identify the numbers in
the text array. As illustrated in the figure 13. (a), number identification first searches for character
“#”. Any character followed by #, is converted to the relevant number. In this case, the text array
consists of English keyboard letters/strings of “AA Amali” font type. Hence, the backward
relation has considered to identify the numbers. In “AA Amali” font use following characters; [c,
w, n, p, o, t, *, ., y, b] to represent 0-9 numbers respectively. The figure 14. (a) shows the
MATLAB code used for sending text array from MATLAB environment to word application.
Figure 14. (b) shows an example where the text array keeps corresponding English
keyboardlettersof “AA Amali” font type in MATLAB environment. After sending it to word
11. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
application, the text array display in Si
completing the Braille to Sinhala translation.
4. RESULTS AND DISCUSSION
In this section, we present results obtained by the developed Optical
system. Many researches claimed that
scanned Braille documents were success
out that the scanned image of
due to less intensity differentiable of the
The first two images in figure 1
documents. As Shown above in our work we were unable to acquire detail images of
documents by scanning method.
Braille dots were written by pen and computer
documents) for the image acquisition (see figure
many scanned images found in internet (figure 1
successfully extracted the Braille
accuracy by regenerating 319/322
figure15(c) respectively.
The performance of the developed
Summaryof the performance evaluation
computer-generated Braille documents were tested in different resolution and different number of
Braille characters. Most of the time
translatedBraille characters to Sinhala or English language
executed on Intel Core i7-7500U CPU @ 2.70GHz
machine. Each tested image in the table 01 executed 10 times
(a)
Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
application, the text array display in Sinhala letters (in AA Amali font type) by successfully
to Sinhala translation.
ISCUSSION
In this section, we present results obtained by the developed Optical Braille
any researches claimed that their work on Braille character recognition using
documents were success [4],[5],[21]–[23]. However, in our case it turn
out that the scanned image of an available Braille documents cannot processed further
less intensity differentiable of the Braille dots with respect to background
images in figure 15(a) acquired by scanning available single side
documents. As Shown above in our work we were unable to acquire detail images of
documents by scanning method. Hence, we initiated with scanned Braille documents where
dots were written by pen and computer-generated Braille documents (screenshots of word
documents) for the image acquisition (see figure 2). Then the developed method was
found in internet (figure 15 (b) and (c)), where the OBT system
Braille characters. The developed OBT system work
/322 and 220/220 Braille characters correctly in figure 1
The performance of the developed OBT system was tested with different input images
Summaryof the performance evaluation is presented in the table 01.Scanned handwritten and
documents were tested in different resolution and different number of
Most of the time, in both types the developed system
characters to Sinhala or English language with 100% accuracy. T
7500U CPU @ 2.70GHz – 2.90GHz, 8GB RAM, Win
Each tested image in the table 01 executed 10 times.
(b) (c)
Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
39
letters (in AA Amali font type) by successfully
Braille Translator
character recognition using
in our case it turned
cannot processed further
s with respect to background.
(a) acquired by scanning available single sidedBraille
documents. As Shown above in our work we were unable to acquire detail images of theBraille
documents where
documents (screenshots of word
was tested with
(b) and (c)), where the OBT system
workedover 99%
in figure 15(b) and
different input images.
Scanned handwritten and
documents were tested in different resolution and different number of
n both types the developed system successfully
The testswere
2.90GHz, 8GB RAM, Windows based
12. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
40
Table 1: Performance of the Optical Braille Translator System
Next, Braille to Sinhala language translation process is discussed. The input Sinhala Braille image
(consists of 66Braille characters)is shown in figure 16(a). The last four Braille characters indicate
a four-digit number as they follow “⠼” Braille character at the beginning of the last row. The
proposed OBT system successfully extracted all 66Braille characters and obtained the binary
equivalent decimal number array which is shown in figure 16(b). In this work our main objective
wasto translate Sinhala Braille document toSinhala language and display the identified Sinhala
characters on a word application. Here, we used “AA Amali” Sinhala font to write text in word
application. Hence, when relating obtained decimal values to Sinhala letters, English keyboard
letters of “AA Amali” font was used in MATLAB environment. The figure 13 (section 3.7)
shows database used for this research work where the English keyboard letters of “AA Amali”
font displayed in green color.
13. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
41
(a) Input Sinhala Braille Image
(b) Corresponding decimal numbers
Accordingly, the figure 16(c) shows the text array corresponds to decimal values. Then the
number identification subroutine was called to identify numbers in the text array.This subroutine
converted the characters in between “#” and next immediate “ ” (space) to numbers. In “AA
Amali” font use following characters; [c, w, n, p, o, t, *, ., y, b] to represent 0-9 numbers
respectively. Accordingly, the output text array is shown in figure 17(d). Then this text array sent
to word application by selecting “AA Amali” as the font type. The word application displayed the
text array in Sinhala langue correctly as “ශ්රඉ ජයව ධන උර වඉශ්වවඉ යආලය භඋතඉක වඉ යආ
අධයනආංශය jraIh 2018”. Since some characters not present in the Sinhala Braille system
with respect to Sinhala language, pronunciation sound of the Sinhala word is used to write word
in the Sinhala Braille system as in above case.
In order to interact with users, a simple graphical user interface (GUI) was developed for the
proposed OBT system.
(c) Text array corresponding to decimal numbers in MATLAB environment
(d) Final output text array MATLAB environment
Figure 16: Sinhala Braille to Sinhala Language Translation in MATLAB
environment
14. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
42
Figure 17: OBT GUI
Here, the user can select image of a Braille document in .jpg,.bmp and .png supporting formats.
The selected image will be visualized on the GUI itself. The current version supports Sinhala
application.
5.CONCLUSIONS
The developed Optical Braille Translator system is capable of translating Sinhala Braille to
Sinhala language or Grade I English Braille to English language over 99% accurately. Further,
OBT successfully recognized numbers in both Sinhala Braille and Grade I English Braille. Some
characters/words in Grade II English Braille and capital letters in both Grade I and Grade II
English Braillecan be translated as well. In conclusion OBT system was successfully
implemented to facilitate communication between visually impaired and sighted persons.
ACKNOWLEDGEMENT
The authors would like to thank the University of Sri Jayewardenepura (Grant No
ASP/01/RE/SCI/2017/13) for the funding support and to Mr. ChanakaGunarathna, Lecturer,
Student Association of Blind, Faculty of Humanities and Social Sciences, University of Sri
Jayewardenepura.
REFERENCES
1. Department of Census and Statistics-Sri Lanka, “Characteristics of the disabled persons census of
population and housing 2001,” 2001. [Online]. Available:
http://www.statistics.gov.lk/pophousat/des_chra.asp. [Accessed: 02-Mar-2016].
2. A. M. S. Al-salman, A. El-zaart, Y. Al-suhaibani, K. Al-hokail, and A. Gumaei, “Designing
Braille Copier Based on Image Processing Techniques,” Int. J. Soft Comput. Eng. ISSN, vol. 4,
no. 5, pp. 62–69, 2014.
3. A. Mousa, H. Hiary, R. Alomari, and L. Alnemer, “Smart Braille System Recognizer,” IJCSI Int.
J. Comput. Sci. Issues, vol. 10, no. 6, pp. 52–60, 2013.
4. S. D. Al-Shamma and S. Fathi, “Arabic braille recognition and transcription into text and voice,”
2010 5th Cairo Int. Biomed. Eng. Conf. CIBEC 2010, pp. 227–231, 2010.
5. J. Li, X. Yan, and D. Zhang, “Optical Braille recognition with haar wavelet features and Support-
Vector Machine,” 2010 Int. Conf. Comput. Mechatronics, Control Electron. Eng. C. 2010, vol. 5,
pp. 64–67, 2010.
6. M. Wajid, M. Waris Abdullah, and O. Farooq, “Imprinted Braille-character pattern recognition
using image processing techniques,” 2011 Int. Conf. Image Inf. Process., no. Iciip, pp. 1–5, 2011.
7. L. Wong, W. Abdulla, and S. Hussmann, “A software algorithm prototype for optical recognition
of embossed braille,” Proc. - Int. Conf. Pattern Recognit., vol. 2, pp. 586–589, 2004.
8. K. P. S. G. Sugirtha and Dhanalakshmi.M, Transliteration of Braille Code into Text in English
Language, vol. 2, no. 2. Springer Singapore, 2018.
9. E. Jacinto Gómez, H. Montiel Ariza, and F. H. Martínez Sarmiento, “There are very few work
previously done for recognizing Sinhala Braille letters.,” Eighth Int. Conf. Graph. Image Process.
(ICGIP 2016), vol. 10225, no. Icgip 2016, p. 102250N, 2017.
10. S. H. Khaled and H. S. Abbas, “Braille Character Recognition Using Associtive Memory,” Int. J.
Eng. Res. Adv. Technol., no. 1, pp. 31–45, 2017.
11. S. Chatterjee, “Creation of an IT Enabled Sinhala to Braille Conversion Engine,” Int. J. Comput.
Appl. Eng. Sci., vol. IV, no. Ii, pp. 17–21, 2014.
12. N. M. T. De Silva and S. R. Liyanage, “Sinhala Braille Translator,” Int. J. Trend Res. Dev., vol. 3,
no. 4, pp. 380–384, 2016.
13. BANA MEMBERS, ENGLISH BRAILLE AMERICAN EDITION. Louisville: American Printing
House for the Blind., 1994.
15. The International Journal of Multimedia & Its Applications (IJMA) Vol.10, No.1/2/3, June 2018
43
14. Braille Authority of North America, “Unified English Braille,” 2016. [Online]. Available:
http://www.brailleauthority.org/ueb.html. [Accessed: 21-Apr-2016].
15. C. Simpson, The Rules of Unified English Braille Edited by, 2nd ed. California: International
Council on English Braille, 2013.
16. MathWorks, “Create COM server - MATLAB actxserver - MathWorks India,” 2016. [Online].
Available:
http://in.mathworks.com/help/matlab/ref/actxserver.html?searchHighlight=actxserver&s_tid=doc_
srchtitle&requestedDomain=in.mathworks.com. [Accessed: 16-May-2016].
17. R. C. Gonzalez, R. E. Woods, and S. L. Eddins, Digital Image Processing. Delhi: Pearson
Education (Singapore) Pte. Ltd, 2004.
18. MathWorks, “Image Processing Toolbox Documentation - MathWorks India,” 2016. [Online].
Available: http://in.mathworks.com/help/images/functionlist.html. [Accessed: 02-Mar-2016].
19. MathWorks, “Radon transform - MATLAB radon - MathWorks India,” 2016. [Online]. Available:
https://in.mathworks.com/help/images/ref/radon.html. [Accessed: 02-Mar-2016].
20. Sinhalese Font, “Aa Amali Font Download ? Free Sinhala Font,” 2016. [Online]. Available:
http://www.sinhalesefont.com/download.php?id=736725. [Accessed: 21-Apr-2016].
21. C. N. R. Kumar and S. Srinath, “A novel and efficient algorithm to recognize any universally
accepted braille characters: A case with kannada language,” Proc. - 2014 5th Int. Conf. Signal
Image Process. ICSIP 2014, pp. 292–296, 2014.
22. G. Morgavi and M. Morando, “A neural network hybrid model for an optical braille recognitor,”
Int. Conf. Signal, Speech …, no. January 2002, 2002.
23. A. Antonacopoulos and D. Bridson, “A robust Braille recognition system,” Doc. Anal. Syst. VI,
pp. 533–545, 2004.
24. Wikipedia, “Bharati Braille,” 2016. [Online]. Available: http://www.bpaindia.org/VIB Chapter-
VI.pdf. [Accessed: 02-Mar-2016].
25. Wiki 2, “Russian Braille chart - Russian Braille — Wikipedia Republished // WIKI 2,” 2016.
[Online]. Available: https://wiki2.org/en/Russian_Braille#/media/File:Russian_Braille_chart.jpg.
[Accessed: 02-Mar-2016].
Authors
Dr. W. K. I. L. Wanniarachchi is working as a Senior Lecturer in the Department
of Physics, Faculty of Applied Sciences, University of Sri Jayewardenepura, Sri
Lanka. He is a graduate in Bachelor of Science (Physics). He received his PhD in
Physics from the Wayne State University, MI, USA. His research interests are on
Computer Vision and Image Processing, Embedded Systems and Electronic
Structure.
T. D. S. H. Perera is working as a teaching assistant in the Department of Physics,
Faculty of Applied Sciences, University of Sri Jayewardenepura, Sri Lanka. He
received B.Sc. degree in Physics from the University of Sri Jayewardenepura, Sri
Lanka in 2017. His research interest includes Image Processing and Optics.