This document summarizes a study that evaluated a new text entry technique called H4-Writer, which was designed for gaze-controlled environments. H4-Writer uses only four directional keys and Huffman coding to map characters to key combinations based on character frequency. Nine participants used H4-Writer with a mouse, gamepad, or eye tracker over three blocks. On average, entry speed was highest with a mouse (3.54 wpm), followed by gamepad (3.33 wpm) and eye tracker (2.11 wpm). Entry speed, efficiency, and KSPC all improved with practice over the blocks as participants learned the key codes. The average KSPC of 2.62 for H
Quick pad tagger an efficient graphical user interface for building annotate...csandit
More and more domain specific applications in the internet make use of Natural Language Processing (NLP) tools (e. g. Information Extraction systems). The output quality of these applications relies on the output quality of the used NLP tools. Often, the quality can be increased by annotating a domain specific corpus. However, annotating a corpus is a time consuming and exhaustive task. To reduce the annotation time we present a custom Graphical
User Interface for different annotation layers.
The lecture presents the open source project Tesseract - a free OCR engine written in C++. The lecture presents the strong and weak sides of tesseract and explains how to train it in a new language. The lecture demonstration materials are available at the authors's blog: http://www.nakov.com/blog
Quick pad tagger an efficient graphical user interface for building annotate...csandit
More and more domain specific applications in the internet make use of Natural Language Processing (NLP) tools (e. g. Information Extraction systems). The output quality of these applications relies on the output quality of the used NLP tools. Often, the quality can be increased by annotating a domain specific corpus. However, annotating a corpus is a time consuming and exhaustive task. To reduce the annotation time we present a custom Graphical
User Interface for different annotation layers.
The lecture presents the open source project Tesseract - a free OCR engine written in C++. The lecture presents the strong and weak sides of tesseract and explains how to train it in a new language. The lecture demonstration materials are available at the authors's blog: http://www.nakov.com/blog
Smart Huffman Compression is a software appliance designed to compress a file in a better way. By functioning as an JSP, it provides high level abstraction of java Servlet. For example, Smart Huffman Compression encodes the digital information using fewer bits, reduces the size of file without loss of data in a single, easy-to-manage software appliance form factor. It also provides us the decompression facility also. Smart Huffman Compression provides our organization with effective solutions to reduce the file size or lossless compression of data. It also expedites security of data using the encoding functionality. It is necessary to analyze the relationship between different methods and put them into a framework to better understand and better exploit the possibilities that compression provides us image compression, data compression, audio compression, video compression etc.
arithmetic and adaptive arithmetic codingAyush Gupta
Data compression techniques in Multimedia and data transmission in digital computers. You can solve the given question on board the solution is on YT. If you search there you'll find it in a few minutes. Hope it'll help you.......
Design and Implementation of LZW Data Compression Algorithmijistjournal
LZW is dictionary based algorithm, which is lossless in nature and incorporated as the standard of the consultative committee on International telegraphy and telephony, which is implemented in this paper. Here, the designed dictionary is based on content addressable memory (CAM) array. Furthermore, the code for each character is available in the dictionary which utilizes less number of bits (5 bits) than its ASCII code. In this paper, LZW data compression algorithm is implemented by finite state machine, thus the text data can be effectively compressed. Accurate simulation results are obtained using Xilinx tools which show an improvement in lossless data compression scheme by reducing storage space to 60.25% and increasing the compression rate by 30.3%.
Performance Improvement Of Bengali Text Compression Using Transliteration And...IJERA Editor
In this paper, we propose a new compression technique based on transliteration of Bengali text to English. Compared to Bengali, English is a less symbolic language. Thus transliteration of Bengali text to English reduces the number of characters to be coded. Huffman coding is well known for producing optimal compression. When Huffman principal is applied on transliterated text significant performance improvement is achieved in terms of decoding speed and space requirement compared to Unicode compression
Text Steganography Using Compression and Random Number GeneratorsEditor IJCATR
A lot of techniques are used to protect and hide information from any unauthorized users such as Steganography and Cryptography. Steganography hides a message inside another message without any suspicion, and Cryptography scrambles a message to conceal its contents. This paper uses a new text steganography that is applicable to work with different languages, the approach, based on the Pseudorandom Number Generation (PRNG), embeds the secret message into a generated Random Cover-text. The output (Stego-Text) is compressed to reduce the size. At the receiver side the reverse of these operations must be carried out to get back the original message. Two secret keys (Hiding Key & Extraction Key) for authentication are used at both ends in order to achieve a high level of security. The model has been applied to different message languages and both encrypted and unencrypted messages. The experimental results show the model‟s capacity and the similarity test values.
Relationship Among Token, Lexeme & PatternBharat Rathore
Relationship among Token, Lexeme and Pattern
Outline
Token
Lexeme
Pattern
Relationship
Tokens : A token is sequence of characters that can be treated
as a unit/single logical entity.
Examples
Keywords
Examples : for, while, if etc.
Identifier
Examples : Variable name, function name, etc.
Operators
Examples : '+', '++', '-' etc.
Separators
Examples : ',' ';' etc.
Pattern
Pattern is a rule describing all those lexemes that can represent a particular token in a source language.
Lexeme
It is a sequence of characters in the source program that is matched by the pattern for a token.
Example : “float”, “=“, “223”, “;”
Proposed Arabic Text Steganography Method Based on New Coding TechniqueIJERA Editor
Steganography is one of the important fields of information security that depend on hiding
secret information in a cover media (video, image, audio, text) such that un authorized person fails to realize its
existence. One of the lossless data compression techniques which are used for a given file that contains many
redundant data is run length encoding (RLE). Sometimes the RLE output will be expanded rather than
compressed, and this is the main problem of RLE. In this paper we will use a new coding method such that its
output will be contains sequence of ones with few zeros, so modified RLE that we proposed in this paper will be
suitable for compression, finally we employ the modified RLE output for stenography purpose that based on
Unicode and non-printed characters to hide the secret information in an Arabic text.
NOTE TAKING IN VR: THE FOREARM KEYBOARDIJCI JOURNAL
This work presents and evaluates a forearm keyboard that allows users to enter textual data using a
natural full-handed typing mechanism for virtual reality head-mounted display environments. Should the
issues noted with the keyboard during the study be solved, the keyboard would compare favourably with
others seen in the literature.
Smart Huffman Compression is a software appliance designed to compress a file in a better way. By functioning as an JSP, it provides high level abstraction of java Servlet. For example, Smart Huffman Compression encodes the digital information using fewer bits, reduces the size of file without loss of data in a single, easy-to-manage software appliance form factor. It also provides us the decompression facility also. Smart Huffman Compression provides our organization with effective solutions to reduce the file size or lossless compression of data. It also expedites security of data using the encoding functionality. It is necessary to analyze the relationship between different methods and put them into a framework to better understand and better exploit the possibilities that compression provides us image compression, data compression, audio compression, video compression etc.
arithmetic and adaptive arithmetic codingAyush Gupta
Data compression techniques in Multimedia and data transmission in digital computers. You can solve the given question on board the solution is on YT. If you search there you'll find it in a few minutes. Hope it'll help you.......
Design and Implementation of LZW Data Compression Algorithmijistjournal
LZW is dictionary based algorithm, which is lossless in nature and incorporated as the standard of the consultative committee on International telegraphy and telephony, which is implemented in this paper. Here, the designed dictionary is based on content addressable memory (CAM) array. Furthermore, the code for each character is available in the dictionary which utilizes less number of bits (5 bits) than its ASCII code. In this paper, LZW data compression algorithm is implemented by finite state machine, thus the text data can be effectively compressed. Accurate simulation results are obtained using Xilinx tools which show an improvement in lossless data compression scheme by reducing storage space to 60.25% and increasing the compression rate by 30.3%.
Performance Improvement Of Bengali Text Compression Using Transliteration And...IJERA Editor
In this paper, we propose a new compression technique based on transliteration of Bengali text to English. Compared to Bengali, English is a less symbolic language. Thus transliteration of Bengali text to English reduces the number of characters to be coded. Huffman coding is well known for producing optimal compression. When Huffman principal is applied on transliterated text significant performance improvement is achieved in terms of decoding speed and space requirement compared to Unicode compression
Text Steganography Using Compression and Random Number GeneratorsEditor IJCATR
A lot of techniques are used to protect and hide information from any unauthorized users such as Steganography and Cryptography. Steganography hides a message inside another message without any suspicion, and Cryptography scrambles a message to conceal its contents. This paper uses a new text steganography that is applicable to work with different languages, the approach, based on the Pseudorandom Number Generation (PRNG), embeds the secret message into a generated Random Cover-text. The output (Stego-Text) is compressed to reduce the size. At the receiver side the reverse of these operations must be carried out to get back the original message. Two secret keys (Hiding Key & Extraction Key) for authentication are used at both ends in order to achieve a high level of security. The model has been applied to different message languages and both encrypted and unencrypted messages. The experimental results show the model‟s capacity and the similarity test values.
Relationship Among Token, Lexeme & PatternBharat Rathore
Relationship among Token, Lexeme and Pattern
Outline
Token
Lexeme
Pattern
Relationship
Tokens : A token is sequence of characters that can be treated
as a unit/single logical entity.
Examples
Keywords
Examples : for, while, if etc.
Identifier
Examples : Variable name, function name, etc.
Operators
Examples : '+', '++', '-' etc.
Separators
Examples : ',' ';' etc.
Pattern
Pattern is a rule describing all those lexemes that can represent a particular token in a source language.
Lexeme
It is a sequence of characters in the source program that is matched by the pattern for a token.
Example : “float”, “=“, “223”, “;”
Proposed Arabic Text Steganography Method Based on New Coding TechniqueIJERA Editor
Steganography is one of the important fields of information security that depend on hiding
secret information in a cover media (video, image, audio, text) such that un authorized person fails to realize its
existence. One of the lossless data compression techniques which are used for a given file that contains many
redundant data is run length encoding (RLE). Sometimes the RLE output will be expanded rather than
compressed, and this is the main problem of RLE. In this paper we will use a new coding method such that its
output will be contains sequence of ones with few zeros, so modified RLE that we proposed in this paper will be
suitable for compression, finally we employ the modified RLE output for stenography purpose that based on
Unicode and non-printed characters to hide the secret information in an Arabic text.
NOTE TAKING IN VR: THE FOREARM KEYBOARDIJCI JOURNAL
This work presents and evaluates a forearm keyboard that allows users to enter textual data using a
natural full-handed typing mechanism for virtual reality head-mounted display environments. Should the
issues noted with the keyboard during the study be solved, the keyboard would compare favourably with
others seen in the literature.
A bidirectional text transcription of braille for odia, hindi, telugu and eng...eSAT Journals
Abstract
Communication gap establishes an aura of unwillingness of understanding the factor behind. Basically, this is one of the common
agenda for the visually challenged people. In order to bridge this gap, a proper platform of learning for both the mass and the
visually challenged for any native language is emphasized in this paper through Braille pattern. Braille, a code, well known to
visually challenged as their mode of communication is now converted into normal text in Odia, Hindi, Telugu and English
languages using Image segmentation as the base criteria with MATLAB as its simulation field so that every person can be able to
easily decode the information being conveyed by these people. The algorithm makes the best use of segmentation, histogram
analysis, pattern recognition, letter arrays, data base generation with testing in software and dumping in using Spartan 3e FPGA
kit. This paper also elaborates on the reverse conversion of native languages and English to Braille making the paper to be more
compatible. The processing speed with efficiency and accuracy defines the effective features of this paper as a successful
approach in both software and hardware.
Keywords: visually challenge, image segmentation, MATLAB, pattern recognition, Spartan 3e FPGA kit, compatible.
HCI BASED APPLICATION FOR PLAYING COMPUTER GAMES | J4RV4I1014Journal For Research
This paper describes a command interface for games based on hand gestures and voice command defined by postures, movement and location. The system uses computer vision requiring no sensors or markers by the user. In voice command the speech recognizer, recognize the input from the user. It stores and passes command to the game, action takes place. We propose a simple architecture for performing real time colour detection and motion tracking using a webcam. The next step is to track the motion of the specified colours and the resulting actions are given as input commands to the system. We specify blue colour for motion tracking and green colour for mouse pointer. The speech recognition is the process of automatically recognizing a certain word spoken by a particular speaker based on individual information included in speech waves. This application will help in reduction in hardware requirement and can be implemented in other electronic devices also.
Braille to text and speech for cecity personseSAT Journals
Abstract
Visually impaired peoples use braille system for reading and writing. This people are not able to operate computers and other educational software due to this they unable to improve their knowledge. They are comparatively weak than people with clear vision which causes adverse effect on the economic society. One way to improve their knowledge is by introducing a system which can convert braille to word and audio which will improve them to communicate freely with outside world. This system presents the implementation of braille to word and audio converter which is output and word of that particular input. Braille Keypad with different combination of six cells are used an input to system. In this system FPGA kit is used to convert this input to English text and also display it on LCD after decoding English text it converted to audio
Keywords: Braille language, blind people, Braille to English word converter.
ISSUES IN IMPLEMENTATION OF PARALLEL PARSING ON MULTI-CORE MACHINESijcseit
The advent of multi-core architecture has highly influenced the area of high performance computing.
Parallel compilation is the area which still needs significant improvement by the use of this architecture.
Recent research has shown some improvement in lexical analysis phase. But it is difficult to implement the
same technique in parsing phase. This paper highlights some issues related to implementation of parallel
parsing on multi-core machines.
ISSUES IN IMPLEMENTATION OF PARALLEL PARSING ON MULTI-CORE MACHINESijcseit
The advent of multi-core architecture has highly influenced the area of high performance computing.
Parallel compilation is the area which still needs significant improvement by the use of this architecture.
Recent research has shown some improvement in lexical analysis phase. But it is difficult to implement the
same technique in parsing phase. This paper highlights some issues related to implementation of parallel
parsing on multi-core machines.
TEXT ADVERTISEMENTS ANALYSIS USING CONVOLUTIONAL NEURAL NETWORKSijdms
In this paper, we describe the developed model of the Convolutional Neural Networks CNN to a
classification of advertisements. The developed method has been tested on both texts (Arabic and Slovak
texts).The advertisements are chosen on a classified advertisements websites as short texts. We evolved a
modified model of the CNN, we have implemented it and developed next modifications. We studied their
influence on the performing activity of the proposed network. The result is a functional model of the
network and its implementation in Java and Python. And analysis of model results using different
parameters for the network and input data. The results on experiments data show that the developed model
of CNN is useful in the domains of Arabic and Slovak short texts, mainly for some classification of
advertisements. This paper gives complete guidelines for authors submitting papers for the AIRCC
Journals.
A Performance Comparison Of C# 2013, Delphi Xe6, And Python 3.4 Languagesijpla
C# 2013, Delphi XE6, and Python 3.4 are the newest and most popular programming languages. These
programming languages become more popular every passing day. In this study, the response times,
memory usages, and code lengths of these languages were tested in various workloads. Whether or not
there was any significant difference between the data obtained from workloads was tested via the Friedman
test. The test indicated a significant difference. In addition, the Wilcoxon signed rank test was used for
determining the effect size. This test showed that the level of the significant difference found in the
Friedman test was high
TEXT ADVERTISEMENTS ANALYSIS USING CONVOLUTIONAL NEURAL NETWORKSijdms
In this paper, we describe the developed model of the Convolutional Neural Networks CNN to a
classification of advertisements. The developed method has been tested on both texts (Arabic and Slovak
texts).The advertisements are chosen on a classified advertisements websites as short texts. We evolved a
modified model of the CNN, we have implemented it and developed next modifications. We studied their
influence on the performing activity of the proposed network. The result is a functional model of the
network and its implementation in Java and Python. And analysis of model results using different
parameters for the network and input data. The results on experiments data show that the developed model
of CNN is useful in the domains of Arabic and Slovak short texts, mainly for some classification of
advertisements.
The advent of multi-core architecture has highly influenced the area of high performance computing.
Parallel compilation is the area which still needs significant improvement by the use of this architecture.
Recent research has shown some improvement in lexical analysis phase. But it is difficult to implement the
same technique in parsing phase. This paper highlights some issues related to implementation of parallel
parsing on multi-core machines.
Correcting optical character recognition result via a novel approachIJICTJOURNAL
Optical character recognition (OCR) is a recognition system used to recognize the substance of a checked picture. This system gives erroneous results, which necessitates a post-treatment, for the sentence correction. In this paper, we proposed a new method for syntactic and semantic correction of sentences it is based on the frequency of two correct words in the sentence and a recursive technique. This approach starts with the frequency calculation of each two words successive in the corpora, the words that have the greatest frequency build a correction center. We found 98% using our approach when we used the noisy channel. Further, we obtained 96% using the same corpus in the same conditions.
Language Identifier for Languages of Pakistan Including Arabic and PersianWaqas Tariq
Language recognizer/identifier/guesser is the basic application used by humans to identify the language of a text document. It takes simply a file as input and after processing its text, decides the language of text document with precision using LIJ-I, LIJ-II and LIJ-III. LIJ-I results in poor accuracy and strengthen with the use of LIJ-II which is further boosted towards a higher level of accuracy with the use of LIJ-III. It also helps in calculating the probability of digrams and the average percentages of accuracy. LIJ-I considers the complete character sets of each language while the LIJ-II considers only the difference. A JAVA based language recognizer is developed and presented in this paper in detail.
Language Identifier for Languages of Pakistan Including Arabic and Persian
H4-Writer
1. H4-Writer: A Text Entry Method Designed For Gaze
Controlled Environment with Low KSPC and Spatial
Footprint
Raeid Saqur
Dept. of Computer Science and Engineering
York University
Toronto, Ontario, Canada M3J 1P3
cs253208@cse.yorku.ca
ABSTRACT
This paper presents a new text entry technique, namely H4-
Writer, designed for gaze controlled environments and
aimed at reducing average KSPC and spatial footprint. It
also presents an empirical evaluation of this proposed
system by using three different input devices: mouse,
gamepad and eye tracker. The experiment was conducted
using 9 participants and the obtained data were used to
compare the entry speeds, efficiency and KSPC of H4-
Writer for all the devices. Over three blocks, the average
entry speed was 3.54 wpm for the mouse, 3.33 wpm for the
gamepad and only 2.11 wpm for the eye tracker. While the
eye tracker fared poorly compared to the mouse and the
gamepad on entry speed, it showed significant improvement
in entry speed over progressing blocks indicating increase
in entry speed with practice had a full longitudinal study
was conducted . The average KSPC of all the three devices
over all the text phrases entered was 2.62, which is
significantly lower compared to other hand writing
recognizing text entry techniques like EdgeWrite. An
analysis of the blocks revealed improvement in error rate,
efficiency and KSPC values with progressing block
numbers as the participants got more acclimatized with the
key codes for corresponding characters.
Author Keywords
Text entry, evaluation, gestures, eye-typing, text input, gaze
control, input devices.
ACM Classification Keywords
H5.2 [Information interfaces and presentation]: User
Interfaces – evaluation/methodology.
INTRODUCTION
Gaze gesture-based text entry techniques using eye trackers
have evolved significantly over the years and with the
increasing availability of eye trackers in size and price
range, the search for more efficient text entry methods that
use eye trackers constitutes a major research area in human
computer interaction.
Text input with eye gaze is and most likely will never be a
mainstream activity [4] due to the inherent limitations
associated with eye-typing [1]. However, gaze controlled
text inputting could be useful to people with severe motor
disability. Also, the idea can be extended to mobile contexts
where people with normal control over their limbs can find
eye gaze interaction useful [4].
On screen keyboards, used for text inputting using eye
gazing, which have one to one mapping of keys to
characters (and commands) usually have large spatial
footprint on the screen due to the large number of keys
used. Other text entry systems include hand writing
recognizing gaze gesture based text entry methods, which
usually have low spatial footprint, however, have high
KSPC (Key Strokes per Character) [7, 12].
The paper proposes an alternative text entry method, H4-
Writer, which was designed keeping gaze controlled
environments in mind. This method addresses the issues
with KSPC and large spatial footprint associated with
conventional text entry methods used for gaze controlled
environments.
The proposed method uses only four directional keys and
hence can be implemented with relatively low spatial
footprint both on screen and using hardware. The method
also uses minimum redundancy codes to map each
character (and command) to the keys based on the
character‟s relative frequency of usage. This ensures
relatively lower KSPC.
This paper describes H4-Writer‟s iterative development
process and an evaluation of the proposed method using
three different input devices.
2. Related Work
The gestural text entry system EyeWrite proposed by
Wobbrock et al. [12] is based on EdgeWrite‟s unistroke
alphabet previously developed for enabling text entry on
PDAs, joysticks, trackballs, and other devices [11] . The
main advantage of H4-Writer over EyeWrite is that H4-
Writer has significantly lower average KSPC (2.32) than
EdgeWrite (3.35) (Figure 1). The KSPC calculation for
these two text entry systems was done by using technique
prescribed by MacKenzie, I. S. [7].
Figure 1: KSPC of English alphabets using EdgeWrite and H4-
Writer.
Other gestural text entry techniques, such as Graffiti [3],
impose problem of text input for people with motor
impairments. Tremor and fatigue dramatically impact a
user‟s ability to make smooth, accurate and controlled
movements [5]. Due to tremor, it is difficult or impossible
for some subjects to make character forms recognizable by
a text entry system like Graffiti.
The proposed H4-Writer addresses this issue because it
does not require the user to use letter-like gestures but
simple dwell time based positioning and clicking gestures.
This is described in more detail in the following section
where the basic implementation of the H4-Writer has been
discussed.
H4-Writer
The name of the proposed gaze-based text entry approach is
derived from the fact that the system uses Huffman coding
and four input keys. As briefly mentioned earlier, one main
advantage of H4-Writer over other similar techniques is that
it uses lower KSPC. This is achieved by deriving unique
prefix, minimum redundancy codes for each letter in the
English alphabet by using Huffman Coding.
The following figure (Figure 2) shows the relative
frequencies of letters in the English language [6].
Figure 2: (a) Relative letter frequencies (b) frequencies in
descending order.
Each alphabet is mapped to its code using these relative
frequencies. So, the alphabets with higher relative
frequencies will have shorter code and vice versa. For
example, the following is a portion of the codes defined for
English alphabets and some commands (Figure 3):
3. Figure 3: Input key codes corresponding to letters and
commands.
Besides the alphabets, codes were derived for „space‟,
symbols and various commands. However, in this paper, we
only consider the lower case English alphabets, space and
other pertinent commands needed for general text entry for
the experiment excluding symbols. As we can see from
Figure 3, the letters with high relative frequency (for e.g.
„e‟) has shorter codes („LL‟) compared to letters with low
relative frequency (for e.g. „q‟ = „LURDRU‟).
An on screen GUI applet containing the four input keys was
developed for evaluating this proposed system. This will be
discussed in more detail in the apparatus section of this
paper.
This paper mainly discusses the pilot research study aimed
at evaluating the performance of the proposed H4-Writer
using different input devices and the skill acquisition trend
among novice users.
METHOD
Participants
Nine participants were recruited from the local university
campus (8 male, 1 female; all of them right handed; mean
age = 23.6 years). All participants were highly experienced
in computer use (average computer usage = 8.6 hours/day),
but none had previously used an eye tracker. All
participants possessed a university degree or were currently
enrolled in university. All subjects rated themselves as
having either good or expert typing skills. All participants
were familiar with using gamepads, however, none of them
had used gamepads for text entry previously. Two
participants reported some experience with Graffiti, a
unistroke handwriting method used by Palm PDAs [3]. Due
to their unfamiliarity with H4-Writer, all the participants
were classified as „novice‟ users.
Apparatus
The experiment was conducted in a quiet lab. Dim lighting
was used while conducting the experiment using eye tracker
and normal room lighting was used for the rest of the
experiment.
Input Devices
This study relied on three different input devices. For text
entry using eye gazing, a non-invasive EyeTech TM3 eye
tracker with a sampling rate of 50 Hz was used (Figure 4).
The resolution was set to 1026 × 768 pixels (14.1 inch LCD
screen).
Figure 4: (a) EyeTech TM3 (b) Laptop setup.
For text entry using gamepad, a Logitech PC gamepad was
used (Figure 5). The four right-hand side numbered buttons
were mapped as the four directional keys used by H4-
Writer.
Figure 5: Logitech PC Gamepad.
For text entry using mouse, a standard size Microsoft
wireless mouse was used (Figure 6).
Figure 6: Microsoft wireless mouse.
4. Application
Figure 7 depicts the H4-Writer application used for the
experiment. The topmost text area displayed the presented
phrase, while the lower one displayed the transcribed text.
The middle section contained the four directional key
buttons, and all the alphabets and commands arranged in
boxes adjacent to the directional keys according to their
corresponding Huffman key codes. The bottom panel was
used to display the results (dependent variables) after each
entry was completed by the user.
The application was designed with the same “look and
feel”. This included audio feedback (“click”) on a key press
event, distinct audio feedback (“beep”) if the wrong
character was entered and a highlight effect on a rollover
event when a key receives focus.
Figure 7: The H4-Writer application used for the experiment.
For example, to transcribe the letter „e‟, the corresponding
code „LR‟ (Figure 3) needs to be keyed. The user does not
need to remember this code. The first key to press is
determined by searching for „e‟ in one of the four boxes
containing the letters, then the key (button) adjacent to the
box is clicked. The first button in this case is „LEFT‟ (code:
„L‟). This step is repeated until the letter (or command) to
transcribe is the only letter left in the boxes. For the case of
„e‟, the second and the last key to press is „RIGHT‟ (code:
„R‟). As soon as the last key is pressed, the letter is entered
in the transcribed text area along with a distinct audio
feedback. Figure 8 depicts this example scenario pictorially.
Figure 8: Entering letter 'e' (a) first key 'L' (b) second key 'R'.
Procedure
Participants first signed an informed consent waiver then
completed a short demographic questionnaire. Next, the
participants were given basic instructions on how to enter
text with each device. Before the participants did the
experiment using the eye tracker, the eye tracker was
calibrated to each participant using twelve calibration
points. Participants were then given five to ten minutes of
unguided practice time to become comfortable typing with
the eye tracker using their assigned condition. For the
experiment, they were instructed to proceed “as quickly and
accurately as possible”.
Participants performed the experiment while sitting down.
They used their dominant hands for the mouse and the
gamepad. For the eye tracker, the distance between the user
and the screen was about 60 cm. Each participant entered
three blocks of phrases for each input device; each block
contained three different randomly chosen phrases from a
500-phrase set [9]. Each study lasted approximately 50
minutes.
5. Figure 9 depicts the basic experimental setup (a) and the
setup for all the three input devices.
Figure 9: Experiment setup (a) basic setup (b) gamepad (c)
mouse (d) eye tracker.
Design
The experiment was a 3 × 3 within-subjects design. There
were two independent variables: input device (mouse,
gamepad, eye tracker) and block (1-3). Each block
contained three phrases of text entry. The dependent
variables were entry speed (wpm), efficiency (%) and
KSPC. Participants were divided into three groups with the
order of input devices counterbalanced to neutralize
learning effects [8]. In total, the number of phrases entered
was 9 participants × 3 input techniques × 3 blocks × 3
phrases/block = 243.
RESULTS AND DISCUSSION
Performance
Figure 10 shows the results for entry speed. On average,
participants entered text at 3.54 wpm (SD = 0.79) using the
mouse, 3.33 wpm (SD = 1.09) using the gamepad and 2.11
wpm (SD = 0.33) using the eye tracker. In relative terms,
the mouse was 67.5% and the gamepad was 57.7% faster
than the eye tracker, which indicates considerable
difference.
Figure 10: Entry speed vs. input device.
The effect on input device on entry speed was statistically
significant (F2,12 = 33.2, p < .0001).
The entry speed results for the eye tracker (2.11 wpm) was
much lower than the eye-typing speeds achieved with
EyeWrite (5 wpm on average with novice users) [12], the
gaze gesture based text entry system discussed in the
„Related Work‟ section of this paper. Some probable
reasons which attributed to this could be that in the
longitudinal study for EyeWrite, the participants were
allowed much longer acclimatization period with the
system, and all the apparatus used for the study were fully
optimized unlike this pilot study of H4-Writer. However,
the learning curve for H4-Writer showed significant
improvement rate of entry speed over longer use. This is
discussed in more detail in a later part of this paper.
The results obtained suggest that using a mouse or a
gamepad for H4-Writer yields significantly higher entry
speed. The reasons for the significant difference is the
directness of interaction using these two devices, and the
interaction limitations associated with the use of eye
trackers and gaze gesture based text entry [1].
Small improvements in the entry speed across the blocks
were observed with all three of the input devices
(Figure 11).
6. Figure 11: Entry speed vs. block.
The effect of block on entry speed was statistically
significant (F2,24 = 33.2, p < .0001), even though there was
a relatively small amount of text entered. Had a full
longitudinal study been conducted, there likely would be
more improvement with practice. Figure 12 depicts the
learning curve over blocks. A trend line was added, based
on the empirical data, projecting the improvement in speed
(wpm) over 5 addition blocks. They indicate a strong
improvement in speed over progressing blocks using all the
three devices; however, more so using the gamepad and the
eye tracker than the mouse. Note the higher R2
values of the
gamepad (R2
= 0.9812) and the eye tracker (R2
= 0.8411)
compared to the mouse (R2
= 0.8145). This is expected,
because the key codes corresponding to the alphabets are
learned and the participants get more acclimatized with the
entry system over progressing blocks.
Figure 12: Text entry speeds with each input device over 3
blocks and added trend line projecting 5 more blocks.
In the following sections, we will discuss the other
dependent variables of the experiment.
Error Rate
Uncorrected Errors
In this experiment, the participants were allowed to correct
any mistakes made while entering the texts if they wanted
to do so. Thus it was possible to make a significant amount
of errors while entering a phrase and yet have the number
of (uncorrected) errors as 0. Hence, although empirical data
for uncorrected errors were collected and recorded, they do
not have statistical significance. For example, the effect of
input device on uncorrected error rate was not statistically
significant (F2,12 = 1.11, p > .05). Also, the effect of block
on uncorrected error rate was not statistically significant
(F2,24 = 0.57, p > .05).
The corrected errors were taken into account while
calculating the efficiency. The following section discusses
efficiency in more detail.
Efficiency
Efficiency was calculated based on the number of corrected
errors. The efficiency decreased with increased number of
corrected errors.
The effect of input device on efficiency was not statistically
significant (F2,24 = 2.96, p > .05). However, the effect of
block on efficiency was somewhat statistically significant
(F2,24 = 6.01, p < .05). This is expected because the number
of corrected errors decreased with progressing block
numbers as the participants got more acclimatized with the
key codes corresponding to the letters and the commands.
Figure 13 depicts the increase in efficiency with
progressing blocks.
Figure 13: Efficiency vs. block.
7. Figure 14 depicts the projected trend in efficiency over five
extra blocks. These show improvement in efficiency over
progressing blocks with all three input devices. Had a full
longitudinal study been conducted, there likely would be
more improvement in efficiency with practice.
Figure 14: Efficiency with each input device over 3 blocks and
added trend line projecting 5 more blocks.
KSPC
The average overall KSPC of all the phrases entered in the
experiment using all the three input devices was 2.63,
which is still significantly lower compared to average
KSPC of EdgeWrite [11].
The effect of input device on KSPC was not statistically
significant (F2,12 = 2.39, p > .05). This is expected because
KSPC is mainly dependent on the key codes used for all the
alphabets. However, the effect of block on KSPC was
somewhat statistically significant (F2,24 = 5.04, p < .05).
Figure 15 shows the average KSPC for each of the input
devices with increasing block numbers. This is expected
because the number of (corrected) errors decreases and
efficiency increases with more practice. Since KSPC is
dependent on the key size, and key size is dependent on the
efficiency and error rate, hence, KSPC decreases slightly
with progressing blocks.
Figure 15: KSPC vs. block.
CONCLUSION
In general, using H4-Writer with mouse and gamepad
yielded better performance and KSPC using just three
blocks. However, eye tracker showed significant
improvement both in terms of performance and KSPC with
progressing blocks as the participants got more
acclimatized with the key codes used in the proposed
system.
This work is important since the new proposed text entry
method is novel and addresses two key issues (KSPC and
spatial footprint) in text entry systems. Future work is
intended to conduct a full longitudinal study of H4-Writer
using fully optimized apparatus.
ACKNOWLEDGEMENT
The author thanks all the participants who volunteered to
take part in this experiment. The author also thanks
Professor I.S. Mackenzie for advising and overseeing all
aspects of this experiment.
REFERENCES
1. Ashmore, M., Duchowski, A. T., and Shoemaker,
Efficient eye pointing with a fisheye lens,
Proc. GI’05 (Victoria: CIPS, 2005), 203-210.
2. Felzer, T. and Nordmann, R., Alternative text
entry using different input methods,
Proc. SIGACCESS Conference on Computers and
Accessibility’06, (New York: ACM, 2006), 10-17.
3. Graffiti. Palm, Inc.
http://www.palm.com/products/input/
4. Isokoski, P., Text input methods for eye trackers
using off-screen targets, Proc. ETRA'00, (New
York: ACM, 2000), 15-21.
5. Keates, S., Hwang, F., Langdon, P., Clarkson, P.
J., and Robinson, P., Cursor measures for motion-
impaired computer users, Proc. Assets '02, (New
York: ACM, 2002), 135-142.
6. Lewand, Robert (2000). Cryptological
Mathematics. The Mathematical Association of
America. p. 36. ISBN 978-0883857199.
7. Mackenzie, I. S., KSPC (keystrokes per character)
as a characteristic of text entry techniques, Proc.
Symposium on Human-Computer Interaction with
Mobile Devices’02, (Heidelberg: Springer-Verlag,
2002), 195-210.
8. 8. MacKenzie, I. S. (2007). Evaluation of text entry
techniques. In MacKenzie, I. S., & Tanaka-Ishii,
K. (Eds.) Text entry systems: Mobility,
accessibility, universality, pp. 75-101. San
Francisco, CA: Morgan Kaufmann.
9. Mackenzie, I. S., and Soukoreff, R. W., Phrase sets
for evaluating text entry techniques, Extended
Abstracts CHI’03, (New York: ACM, 2003), 754-
755.
10. Wobbrock, J., The benefits of physical edges in
gesture-making: empirical support for an edge-
based unistroke alphabet, Extended Abstracts CHI
'03, (New York: ACM, 2003), 942-943.
11. Wobbrock, J. O., Myers, B. A., and Kembel, J. A.,
EdgeWrite: a stylus-based text entry method
designed for high accuracy and stability of motion.
Proc. UIST '03, (New York: ACM, 2003), 61-70.
12. Wobbrock, J. O., Rubinstein, J., Sawyer, M. W.,
and Duchowski, A. T., Longitudinal evaluation of
discrete consecutive gaze gestures for text entry.
Proc. ETRA '08, (New York: ACM, 2008), 11-18.