This Presentation is on Character Recognition using Artificial Neural networks,
Presented to
Farhana Afrin Duty
Assistant Professor
Department of Statistics
Jahangirnagar University
Savar, Dhaka-1342, Bangladesh
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
Character Recognition using Data Mining Technique (Artificial Neural Network)
1. This Presentation is
on
Character Recognition using Artificial Neural Network
Sudipto Krishna Dutta
20204021
4th Batch
February 24, 2022
Presented to
Farhana Afrin Duty
Assistant Professor
Department of Statistics
Jahangirnagar University
Savar, Dhaka-1342, Bangladesh
2. Contents
• Introduction
• Problem Definition
• Classification of Character Recognition
• Image Processing
• Design for the Artificial Neural Network (ANN) & Methodology
• Literature Survey
• Proposed Methodology & Results
• Conclusion
• Improvement & Future Work
• References
3. Introduction
• Data mining is a powerful method and it’s techniques are very
useful for character recognition from different patterns or
images and textual data sets also.
• Techniques that performs Handwritten recognition is able to
detect and acquire characters from paper documents, images
and other sources.
• There exists various data mining techniques which are used for
Handwritten Character Recognition, from them Artificial Neural
Network (ANN) is a great successor for doing this job.
5. Problem definition (continue)
• Current Approaches
OCR – What it is?
Optical Character Recognition or OCR is a technology that enables us
to convert different types of documents, such as scanned paper
document, PDF files or images captured by a digital camera into
editable and searchable data.
• But, OCR isn’t a complete solution/success. So, there are some
solutions given below,
Apply new features extraction Technique
Using Data Mining Techniques (Artificial Neural Network)
Expected 100% accuracy of character identification.
6. Classification of Character Recognition
• CRS consists of various features such as: Geometric blur, Spin image, Patch descriptors. Shape
contexts, feature transform and required filters. These features are extracted at training time of the
system.
• Features are selected to learn maximum knowledge about the given characters to achieve high
accuracy.
• There are various Machine Learning techniques such as KNN, random forest, support vector
machine and CNN are used for extraction of text from various scenes.
• The actual training of the machine is to provide Artificial Intelligence (AI) which automatically learn
and improve efficiency, and is called Machine learning.
• The machine is focused on developing computer programs to access and use data. Artificial Neural
Networks (ANN) is learning model that is used in machine learning.
• All knowledge that is acquired by animal or human neural network is simulated by Artificial
Neural Networks (ANN).
• Benefit of this learning model is that they are very useful in automating the tasks in which the
decision of a human being is too long or humiliated.
7. Image Processing
• Image Processing is a technique is used to
manipulate the digitized images to extract
meaningful knowledge from it.
• Image processing contains following steps shown in
this right side Figure.
Firstly handwritten document was digitized.
After that preprocessing, segmentation and feature
extraction is performed.
After feature extraction classification is performed
using machine learning techniques along with data
mining technique to recognizing character .
8. Design for the Artificial Neural Network (ANN)
&
Methodology
Collect features from characters
(Using geometry based character features)
Output
Measure the accuracy of the recognition
Artificial
Neural
Network
Pre Processing
Segmentation
Feature Extraction
Classification and
Recognition
Features
value
9. Literature Survey
• Available literature conveys that various approaches have been made in order to
accomplish the task of character recognition.
• In about all the soft computing approaches Artificial Neural Network (ANN) has
been a backend of character classification. This is due to its faster computation.
• The methods used in front end could be (a) statistical approaches (b) kernel
methods (c) support methods or (d) hybrid of fuzzy logic controllers.
• A detailed analysis of some methods is given in Table below which shows the
References, approach and its corresponding accuracy.
10. Proposed Methodology & Results
• It should be mentioned that the proposed algorithm is currently under the rigorous testing
by the authors.
• However ,they have tested the developed algorithm with 5 capital letters and 5 small English
character alphabets. The result is given in Table below.
11. Results (Continue)
• Two main patterns decide the performance of the algorithm,
Success Rate
False Rejection Rate (FRR)
• The success rate defines the rate of recognition whereas the FRR defines the ratio
between the unrecognized patterns and the total number of testing patterns.
• This plot below shows the relative success rate versus the false rejection of the
tested patterns.
12. Conclusion
• This work proposed an algorithm for Hand Written English
alphabet pattern recognition.
• The algorithm is based on principle of Artificial Neural
Network (ANN). However, the algorithm was tested with 5
capital letters set and 5 small letters set.
• The result and discussion section showed that the developed
algorithm works well with maximum recognition rate if
92.59% and minimum False Rejection Rate (FRR) of 0%.
13. Improvement & Future Work
• Make the system more font independent.
• Try different recognition algorithm such HMM(Hidden Markov
Model).
• Improve the efficiency of the feature extraction method.
• Improve the system to indentify any other characters.
14. References
[1] S. e. a. Basu, "Handwritten 'Bangla' alphabet recognition using an MLP based classifier," Proceding of 2nd National Conference on
Computer Processing of Bangla,pp.285-291, Dhaka, 2005.
[2] A. a. S. D. Pal, "Handwritten English Character Recognition Using Neural Network," International Journal of Comuter Science and
Communication, Vol.1, No.2, pp. 141-144, U. P. Technical University, Lucknow, India, 2010.
[3] A. U. e. a. Dinesh, "Isolated handwritten kannada numeral recognition using structural feature and K-means cluster," IISN, pp. 125-129,
India, 2007.
[4] Y. a. C. A. Perwej, "Neural Network for Handwritten English Alphabet Recognition," International Journal of Computer
Application, Vol. 20, No.7, PP. 1-5, India, 2011.
[5] A. K. J. T. T. O. D. Trier, "Feature extraction methods for character recognition- A survey," Pattern Recognition, Vol. 29, No.4, pp. 641-662,
Chicago, 1996.
[6] H. a. F. S. A. Imtiaz, "A wavelet-domain local dominant feature selection scheme for face recognition," International Journal of Computing
and Business Research, Vol.3, Issue.2, India, 2011.
[7] P. J. e. al, "Diagonal based feature extraction for handwritten alphanbet recognition system using neural network," International Jouranla of
Computer Science and Information Technology, Vol. 3, No. 1, pp.27-38, India, 2011.
[8] C. a. B. K. Kurian, "Continious speech recognition system for Malayamlam langauge using PLP cepstral coefficient," Journal of Computing
and Business Research, Vol.3, Issue.1, London, 2012.
[9] S. a. D. S. Sivanandam, "Principles of Soft Computing," Wiley- India Publisher, 2nd edition, India, 2011.
[10] S. e. a. Impedovo, "Optical character recognition," International Journal of Pattern Recognition and Artificial Intelligence, Vol. 5(1-2), pp.
1-24, Singapore, 1991.
[11] S. e. a. Mori, "Historical review of OCR research and development," Proceedings of IEEE, vol. 80, pp. 1029-1058, Manhattan, New York,
U.S, 1992.
[12] a. A. S. N. Babu, "Character recognition in historical handwritten documents–a survey," international conference on communication and
signal processing (ICCSP). IEEE, pp. 0299-0304, Barcelona, Spain, 2019.
[13] H. T. S. H. A. M. D. Al-Malah, "Cloud Computing and its Impact on Online Education," IOP Conference Series: Materials Science and
Engineering, vol. 1094, pp. 012024, Aceh, Indonesia, 2021.
[14] B. M. C. M. D. A. a. M. J. S. B. L. Von Ahn, "recaptcha: Human-based character recognition via web security measures," vol. 321, no. 5895,
pp. 1465-1468, London, 2008.
[15] B. A.-K. M. Mahmood, "Review of neural networks and particle swarm optimization contribution in intrusion detection," Periodicals of
Engineering and Natural Sciences (PEN), vol. 7, no. 3, pp. 1067-1073, Bosnia and Herzegovina, 2019.
[16] E. K. A. H. T. H. H. A. a. K. N. N. S. Alseelawi, "Design and Implementation of an E-learning Platform Using N-TIER Architecture,"
international Journal of Interactive Mobile Technologies, vol. 14, no. 6, pp. 171-185, Germany, 2020.
15. All those things concludes my presentation
Thank you for patience hearing
Editor's Notes
What do you mean by Multilayer Perceptron?
A multilayer perceptron (MLP) is a class of feedforward artificial neural network (ANN). ... An MLP consists of at least three layers of nodes: an input layer, a hidden layer and an output layer. Except for the input nodes, each node is a neuron that uses a nonlinear activation function.
What is multilayer perceptron example?
A fully connected multi-layer neural network is called a Multilayer Perceptron (MLP). If it has more than 1 hidden layer, it is called a deep ANN. An MLP is a typical example of a feedforward artificial neural network.
What is multilayer perceptron in neural network?
A multilayer perceptron (MLP) is a class of feedforward artificial neural network (ANN). ... An MLP consists of at least three layers of nodes: an input layer, a hidden layer and an output layer. Except for the input nodes, each node is a neuron that uses a nonlinear activation function.