PATTERNS IDENTIFICATION FOR HITTING ADJACENT KEY

ERRORS CORRECTION USING NEURAL NETWORK MODELS

Jun Li

Department of Land Economy, University of Cambridge, 19 Silver Street, Cambridge, U.K.

Karim Ouazzane, Sajid Afzal, Hassan Kazemian

Faculty of Computing, London Metropolitan University, London, U.K.

Keywords: QWERTY Keyboard, Probabilistic Neural Network, Backpropagation, Key Distance, Time Gap, Error

Margin Distance.

Abstract: People with Parkinson diseases or motor disability miss-stroke keys. It appears that keyboard layout, key

distance, time gap are affecting this group of people’s typing performance. This paper studies these features

based on neural network learning algorithms to identify the typing patterns, further to correct the typing

mistakes. A specific user typing performance, i.e. Hitting Adjacent Key Errors, is simulated to pilot this

research. In this paper, a Time Gap and a Prediction using Time Gap model based on BackPropagation

Neural Network, and a Distance, Angle and Time Gap model based on the use of Probabilistic Neural

Network are developed respectively for this particular behaviour. Results demonstrate a high performance

of the designed model, about 70% of all tests score above Basic Correction Rate, and simulation also shows

a very unstable trend of user’s ‘Hitting Adjacent Key Errors’ behaviour with this specific datasets.

1 INTRODUCTION

Computer users with disabilities or some elderly

people may have difficulties in accurately

manipulating the QWERTY keyboard. For example,

motor disability can cause significant typing

mistakes. These issues haven’t been well addressed

by current technologies. Although alternative input

devices or software such as keyguard and Dasher are

available for use, none of them prove more efficient

or comfortable than the conventional QWERTY

keyboard. Some efforts associated with standard

keyboard has been made such as Windows’

Accessibility Options, ProtoType, however the

solution to remedy typing difficulties encountered

by disabled people hasn’t been achieved as yet.

Ouazzane and Li (2008) provide a

comprehensive report and classified user

performance as four categories, i.e. Motor disability,

Dyslexia, Unfamiliar with Computer and Others

performance. The ‘Miss-stroke’ or ‘Press Additional

Keys’ error is classified as sub-category three within

category one. However, in the paper only a brief

model framework is given based on multi-

technologies to tackle the typing errors as a whole,

while there is no specific solution and convincing

results for solving the ‘Miss-stroke’ error.

In the field of computer science, a neural

network is a mathematical model or computational

model that is inspired by the structure and/or

functional aspects of biological neural networks.

Modern neural networks are non-linear statistical

data modeling tools. They are usually used to model

complex relationships between inputs and outputs or

to find patterns in data. There are three major

learning paradigms, i.e. supervised learning,

unsupervised learning and reinforcement learning,

while BackPropagation is a supervised learning

method and most popular used in practice. It is an

implementation of the Delta rule, whose architecture

belongs to feedforward network. Probabilistic

Neural Networks (PNN) is a type of radial basis

network suitable for classification problems. It is a

feedforward network built with three layers, and

offers a series of advantages, including rapid

training speed, incremental training, and robustness

to noisy examples.

Li J., Ouazzane K., Afzal S. and Kazemian H..

PATTERNS IDENTIFICATION FOR HITTING ADJACENT KEY ERRORS CORRECTION USING NEURAL NETWORK MODELS.

DOI: 10.5220/0003413700050012

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 5-12

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

In the following sections, a Time Gap model and

a Prediction using Time Gap model based on

BackPropagation are developed respectively to

verify the influence of parameters such as keyboard

layout, key distance and time gap on human typing

behaviour. Then, an innovative model named

Distance, Angle and Time Gap model based on

Probabilistic Neural Network is developed to

simulate and predict a specific user typing behaviour

– ‘Hitting Adjacent Key Errors’ as a pilot research.

2 TIME GAP MODELLING

From Fitts’law (Fitts, 1954), users input

performance IP in bits per second is proportional to

the variable movement time ID, which has a direct

relation with the moving distance from one point to

another. Let’s consider a standard keyboard layout,

the time gap between two consecutive strokes

directly depends upon the distance between those

two keys. As observed, the last key’s position

represented by the distance and angle with the target

typing key could affect some of the disabled users’

judgment on their typing accuracy and speed, which

would be reflected by the time gap recorded on the

computer log. Given the user’s typing history, a 1-

gram neural network model named as Time Gap

Neural Network (TGNN) is designed here to

simulate and predict the two consecutive typing

letters’ time gap. A typical structure of generated log

is shown below.

Figure 1: Example structure of a generated log.

It is extracted from a charity helpline keystroke

log. The associated computer is routinely used as a

question recording, database query and email writing

tool by a disabled volunteer. From the reflected

keystroke log, the typing mistakes are predominantly

about adjacent key press errors. The keystroke

recording tool used in this research is KeyCapture

software(Soukoreff and MacKenzie, 2009), which

has been modified and adjusted for the purpose of

this research. It runs in background under Windows

environment to collect keystrokes without

interfering with user’s work.

A function - OnBnClickedsuggesttimegap is

programmed to pre-process the dataset. A fifty-four

virtual key codes set is considered, which includes

fifty-three visible symbols such as alphabet,

numbers and space. The other symbols are classified

as an assumed symbol - ‘Other’.

OnBnClickedsuggesttimegap function only

extracts the keystrokes whose time gaps is in a range

of [0, 3000] ms. The rest keystrokes which have

been considered as either out of range or computer

system related problems are ignored. 2-gram dataset

is created with their corresponding time gaps. This

requires 108 (i.e. NumberOfSymbols * Gram)

neurons in the input layer. All the time gap values

are normalized into a range of [-1, 1] according to

Min-Max Normalization before they are used by the

Time Gap Neural Network model (TGNN). The

normalization equation is shown below,

min max min max min min

' = ( )*( ' ' )/( ) + 'vvV V V VV V

−

−−

(1)

Where

max min

'1,' 1VV

=−

and variable v is the

time gap value extracted from the dataset. The

results of TGNN model will be reversed to their

natural values based on the same equation.

Then, a traditional BackPropagation neural

network is designed with a 108-7-1 three layer

structure. The input includes two consecutive

symbols represented by unary codes, and the output

is the expected time gap between these two

consecutive symbols. MATLAB neural network

toolbox is used for programming. The ‘tansig’ and

‘purelin’ functions are considered as the hidden and

output layer’s activation function.

A reconstructed dataset extracted from the log

file is used as neural network’s training dataset;

another two datasets, i.e. English alphabets in an

alphabetical order and QWERTY keyboard layout

order respectively, ‘abcdefghijklmnopqrstuvwxyz’

and ‘qwertyuiopasdfghjklzxcvbnm’, are used as two

testing cases. The experimental results generated by

TGNN model then demonstrated on these two

datasets are shown in Figure 2 and 3.

Firstly, the TGNN (Time Gap Neural Network)

model is trained based on the log file. Then the

Alphabet and QWERTY sequences are applied to

the TGNN model. Figure 2 shows a simulation of

the user’s typing behaviour (e.g. speed and time gap)

by typing an Alphabet sequence; Figure 3 shows a

simulation of the user’s typing behaviours (e.g.

speed and time gap) by typing a QWERTY

sequence. Due to no predecessors, both

corresponding time gaps of the first keystrokes in

sequence (in Figure 2 is ‘a’; and in Figure 3 is ‘q’)

are counted as zero.

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

Figure 2: Modelling time gap using AÎZ sequence.

Figure 3: Modelling time gap using QWERTY sequence.

In Figure 2 and 3, x-axis represents user’s typing

sequence; y-axis represents the time gap in

milliseconds. Between each two consecutive

alphabets, a blue line is drawn to illustrate the

elapsed time. The maximum time gap (637.4ms)

occurs in Figure 2 when the finger moves from key

‘x’ to ‘y’; while the minimum time gap (89.9ms)

appears in both figures, when the finger moves from

‘j’ to ‘k’.

These two figures show that the current

keystroke’s predecessor affects the user’s typing

behaviour (e.g. time gap) if one ignores the user’s

keystroke action itself and behaviour randomicity

that human may have. Due to the distance difference

between each two keys in computer QWERTY

keyboard, the time gap of each two consecutive keys

during user strokes varies.

The red lines in Figure 2 and 3 represent the

average time cost of all twenty-five movements,

which show that the cost of typing an alphabet order

sequence is 384.44ms (see Figure 2), whereas the

cost of typing a QWERTY order sequence is

342.50ms (see Figure 3). The test shows typing an

Alphabet sequence is more time consuming based on

a standard keyboard. This can be explained by

movement cost, meaning that an alphabet order

sequence would require more time for a user to

locate the keys from one to another.

This research gives a glance at the idea that the

Time Gap between two consecutive keystrokes is

influenced by current symbol’s predecessor. A

further research tracing back more than one gram

history accompanied with a larger dataset is

necessary.

3 PREDICTION USING TIME

GAP

As mentioned in the introduction, people with motor

disability or Parkinson disease using keyboard may

press adjacent keys or stick keys. These can be

shown from the time gap between each two

consecutive key strokes. For example, a time gap

between the windows keyboard messages caused by

sticking keys can be much smaller than the user’s

normal typing speed; the opposite case may also

happen when more time can be spent by disabled

people aiming at the target before making up their

mind. From observation, interestingly it is rare for

those people to completely miss typing a symbol.

According to these distinct behaviours, a neural

network model using BackPropagation (newff) is

the input layer, called Prediction using Time Gap

(PTG). Here, a small sample typed by a Parkinson

person is used to demonstrate the idea. The target

typing sample is,

the quick brown fog jumped over the lazy dog

The user’s true typing sample is,

hthe quick brrooownn fgow jummppefd

iobverethe lwqazy dooggfg

The typed sample is reconstructed for

preprocessing,

@the quick br@o@@wn@ @@f@ox@

jum@p@e@d @@o@ver the l@@azy

do@g@@@

Where the symbol ‘@’ represents an error or a

NULL, compared to the right sample which should

be recognized by PTG model. During preprocessing,

the time gap value which is one of the input

PATTERNS IDENTIFICATION FOR HITTING ADJACENT KEY ERRORS CORRECTION USING NEURAL

NETWORK MODELS

parameters is categorized into three levels and

converted into three bits unary codes. In this case,

‘<= 10 milliseconds’ over-fast => 001

‘10< && <=1000 milliseconds’ user-Speed => 010

‘>1000 milliseconds’ over-slow => 100

The user’s typing has been recorded both by

Notepad and KeyCapture software.

Prediction using Time Gap model is designed

with three layers 30-7-28 structure, where the input

requirement of PTG model is twenty seven length

unary coding symbol {‘a’…’z’, space} and three

length unary coding time gap, and the output

requirement is twenty eight length unary coding

limited in symbol set {‘a’…’z’, space, ‘@’}, where

the symbol ‘@’ is added to represent an additional

or missed symbol.

The correction rate distribution within one

hundred times training is shown in Figure 4, which

has a mean value of 0.8480 and a deviation of

0.0501. The x-axis represents the correction rate

based on the comparison between the target dataset

and PTG generating dataset; the y-axis represents

the absolute frequency of the one hundred times

training results, which illustrates the number of

times a particular outcome occurs.

Figure 4 demonstrates the range that PTG

model‘s correction rate lies on. It shows that the

results lie predominantly between 65% and 90%.

Under this test sample there is about twenty-seven

times where the correction rate has reached near

90% and only once the correction rate happens to be

less than 65%.

This test indicates that the time gap can be

considered as an input element used by neural

network model to correct wrong typed symbols. Due

to no gram consideration and the size limitation of

training dataset, the relationship built between input

and output is a pure right-wrong relationship. This

could lead to a further research on the n-gram

language modelling with larger training and testing

dataset.

4 PROBABILISTIC NEURAL

NETWORK MODELLING

♦ Assumption: the research carried out in

this section is based on one finger typing

user case. User’s each key press and move

rely entirely on a single finger. Skilful

users’ typing behaviour in controlling

Figure 4: Absolute Frequency of PTG model Correction

Rate.

fingers may vary, and the distance of

fingers move between two consecutive

keystrokes could be more complex.

♦ Key Distance Definition: according to the

layout of a computer QWERTY keyboard,

there exists a physical distance between

each two keys. Let

,ij

d be the distance

between key i and key j, and define the

measure unit as key-distance. Then,

shows that the distance between

key ‘A’ and key ‘S’ is one key-distance;

means there are three key-

distances between key ’A’ and key ‘F’.

Users move their fingers toward the next

key as soon as they finish current key

press. The distance between two keys

affects a user’s typing performance.

♦ Error Margin Distance (EMD)

Definition: based on Key Distance, a

variable

dΔ is further defined as a

distance between a user’s typed key -

key and target key -

key and called

Error Margin Distance. The Error Margin

Distance is mainly caused by the user’s

‘Hitting Adjacent Key Error’.

♦ Key Distance Class Definition: let’s

define a class,

{|}

key j ij i

C key key= , by

giving

, { ,..., }

iij n

key key key key∈ ,

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

where

,ij n≤ , n is the number of keys

related to a computer QWERTY keyboard,

key represents a key set around

key

within j key-distances. For instance, a one

key-distance set corresponding to key ‘s’

is,

,1 1

{ | } {' ',' ',' ',' ',' ',' '}

CssDEWAZX=≈

Noisy data prediction models such as Focused Time-

Delay Neural Network not only can be generally

used to analyze a language text, but also can be

explored to analyze some specific problems. For

example, let’s take the helpline data as a real

scenario. As shown in the data, a typist is frequently

making ‘Hitting Adjacent Key Errors’ mistakes.

Therefore, all the typing mistakes are extracted from

the log file and used to identify the possible rules. A

sample of it is shown below,

Figure 5: A sample of ‘Hitting Adjacent Key Errors’.

This is a typical ‘Hitting Adjacent Key Errors’

typing mistake that occurred within a user’s typing

stream. The user’s intention is to type a letter ‘d’

following letter ‘q’, but the letter ‘s’ is mistakenly

pressed. So the user has to go back and make a

correction by pressing ‘backspace’ key shortly after

the mistake is made (in virtual key code, the

‘backspace’ is represented by ‘BACK’). Both Key

Distance and Time Gap are calculated and recorded

in the log.

The user investigation shows users’ Hitting

Adjacent Key behaviour is related to the positions of

both the last key and the current key if one ignores

the stroke randomicity that users’ symptoms may

cause. It also shows that a user’s typing speed

moving from one key to another also plays an

important role in making such errors. For example,

although a faster typing speed than a user’s normal

speed increases the occurrence of ‘Hitting Adjacent

Key Errors’, the users’ hesitation which leads to

much slower typing speed does not always help to

an increase of right typing rate, as shown in the log

file.

Here, the idea is to use these essential

parameters, namely, Key Distance, Time Gap and

Error Margin Distance to discover the fundamental

rules behind uses’ typing mistakes. Let’s start with

the introduction of QWERTY keyboard layout, and

consider Figure 6 and 7,

Figure 6: A QWERTY keyboard layout sample.

Figure 7: Relationship – angle between keys and its

surrounding keys D, E, A.

In Figure 6, key ‘S’ is surrounded by one key-

distance dataset {‘W’, ‘E’, ‘A’, ‘D’, ‘Z’, ‘X’} and two

key-distance dataset {‘Q’, ‘R’, ‘caps lock’, ‘F’, ‘|’,

‘C’}. Given certain inputs, if one requires the neural

network model to be able to produce the right

symbol that a user intends to type, the designed

model not only need to deduce the dataset which the

right symbol belongs to, but also the right angle the

user intends to move towards. This is illustrated in

Figure 7. All keys surrounding ‘S’ are positioned

with different angles. Let’s assume the circle starts

from right-hand side of ‘S’ and turns in an

anticlockwise direction. Then the key ‘D’ can be

expressed by a three dimensions vector,

key =

{key=’S’, distance=1, angle=0}, where key=’S’

illustrates the dataset surrounding key ‘S’,

distance=1 & angle =0 represent the key which is

one key-distance away from key ‘S’ with an angle of

zero degree. The key ‘A’ can be expressed as

key

= {key=’s’, distance=1, angle=π}, distance=1,

angle=π means the key is one key-distance away

from key ‘S’ with an angle of π degree.

The key distance and time gap between last two

grams could determine the error margin between the

wrong key and the right key. In order to prove this

PATTERNS IDENTIFICATION FOR HITTING ADJACENT KEY ERRORS CORRECTION USING NEURAL

NETWORK MODELS

hypothesis, a Neural Network Topology with

Distance, Angle and Time Gap vectors in the input

layer, and the Error Margin Distance vector between

the typed key and target key in the output layer is

designed. These require a precise measurement on

both input and output parameters. However, given

the difficulty of QWERTY keyboard and its

associated operating system to respond to an

accurate simulation of users’ movement and the

difficulty of a neural network to provide a precise

output, this solution, as it stands, is not practical.

For example, the difference in angle between key ‘S’

Æ key ‘E’ and key ‘S’ Æ key ‘R’ is not significant.

This high precision requirement raises the design

difficulty of a neural network model.

In order to overcome these obstacles, a more

robust neural network model with re-designed

vectors on both input and output layers is developed

in this research. The input of neural network model

uses (x, y) coordinate expression instead of distance

and angle, where x represents x-axis key-distance

(i.e. horizontal distance), and y represents y-axis

key-distance (i.e. vertical distance). X-axis key-

distance refers to a user’s horizontal move toward

the typed key; y-axis key-distance refers to a user’s

vertical move toward the typed key. The time gap

parameter is kept unchanged, which represents the

time difference (ms) between two consecutive key

strokes. When the error margin is calculated, the

coordinate centre lies at the current typed key. When

the distance of last typed key and current typed key

is calculated, the coordinate centre lies at the last

typed key. The sign of key distance will be

determined as soon as the coordinate centre is fixed.

In QWERTY keyboard there are maximum of

six one key-distance keys around each key. The user

investigation records suggest that most of ‘Hitting

Adjacent Key Errors’ occur in an area where the

keys are equal or less than one key-distance away

from the target keys. Therefore, instead of

computing a precise error margin

,tf

dΔ

, the output

of neural network model can be designed as a six-

classes classifier. If one counts the class in a wise-

clock direction according to traditional coordinate,

then, from Figure 7, ‘d’ belongs to class one, ‘e’

belongs to class two and so on. Thus the question

can be interpreted as finding an appropriate neural

network model to solve a classification issue

associated with input vectors: Distance, Angle and

Time Gap.

It is well known that radial basis networks can

require more neurons than standard feedforward

BackPropagation networks, but quite often they can

be designed in a fraction of the time it takes to train

standard feedforward networks. One of Radial basis

networks is Probabilistic Neural Networks (PNN)

which can be used for classification problems. As

PNN is a time-efficient and classification-solving

solution, in this research a 3-N-1 structure model,

i.e. Distance, Angle and Time Gap PNN model

(DATP model) is designed based on PNN to predict

where the target key could possibly lie against the

wrong key press.

The DATP model consists of three layers, input

layer, hidden layer and output layer. The hidden

layer – radbas layer compute the distance between

the input vector and the hidden weights vector, and

then produces a distance vector which indicates how

close the input is against the correct letter. The third

layer would classify the results of radbas layer and

produces the right class.

In this experiment, thirty three ‘Hitting Adjacent

Key Errors’ are identified from log file, and are

converted into the format training dataset manually.

At the same time another ten samples are used as

test samples. Here an example is given to show the

pre-processing procedure,

Figure 8: An example of pre-processing procedure.

The first four lines are extracted from log file.

The line following an arrow is the data transformed

manually from the lines above, which has four

parameters, namely, horizontal distance, vertical

distance, time gap between two consecutive

keystroke, and class.

The first line shows that the horizontal distance

from ‘C’ to ‘J’ is 3.5 key-distances, however, if the

move are from ‘J’ to ‘C’, the key-distance would be

-3.5; the vertical distance is one key-distance; the

time gap from ‘C’ to ‘J’ is 108ms (shown in red) and

the class is ‘4’ as the key ‘H’ is at the left hand side

of key ‘J’. In the case of overlapping keys, a half

key-distance can be counted. For example,

Figure 9: An example of overlapping keys pre-processing.

This is a typical key press with overlapped key

‘G’ and key ‘H’. The time gap between ‘G’ press

and ‘H’ press is 3ms, which is much less than the

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

user’s usual typing speed. This has been proved by

the user’s correction which happened afterwards, as

shown in the log file. The horizontal key-distance

between key ‘D’ and key ‘G’ is two key-distances,

however, another 0.5 key-distance is added in pre-

processing by taking into consideration the

overlapping. The vertical distance between these two

keys is zero, while the time gap is 218ms and the

output class is 4.

The experimental results show a correction rate

of 50% which is five out of the ten testing samples.

However, due to the highness of user’s typing

disorder and the small size of training dataset, a

random training and testing dataset selection

strategy is further adopted. The thirty three training

samples and ten testing samples are mixed up and

the random function iRand is applied to randomly

pick up the training dataset and testing dataset in a

proportion of 2/3 and 1/3 respectively. Two groups

of trials are carried out, and each group of them

includes ten training and testing samples. The

corresponding plots are shown in Figure 10.

The x-axis refers to training and testing samples

that are picked up randomly; the y-axis refers to the

prediction rate of the DATP model. The dashed line

in red shows the prediction rate of each testing

dataset according to its training dataset; the line in

blue is the random prediction rate which has been

named as Basic Rate.

The first plot of Figure 10 demonstrates that

there are six rounds out of eight whose prediction

rates are above Basic Rate, while the rest are below

Basic Rate. The highest score (40%) occurs at the

third round, while the lowest score occurs at eighth

round (0%). The second plot indicates that there are

seven rounds whose prediction rates are above Basis

Rate, while the three remaining rounds are below

Basic Rate. The highest score (36%) occurs at the

tenth round while the lowest score (7%) occurs at

the third round.

Both plots show that there are 70% of all tests

scoring above Basic Rate. They also demonstrate a

very unstable trend of user’s ‘Hitting Adjacent Key

Errors’ behaviour. It recommends that the training

dataset with a small size of data may not be able to

give a high prediction rate as the dataset has a bad

convergence. In that case, several rounds of training

with a random dataset selection strategy may be

required.

Further work to be carried out should focus on two

areas: the DATP model development with larger

scaled data to obtain a more accurate prediction rate,

and a touch keyboard combining the sensitivity of

Figure 10: Hitting Adjacent Key prediction rates based on

PPN network (the top and the beneath plots are generated

from group one and two trials respectively).

touch screen and functionality of QWERTY layout

to detect the users’ finger movement more precisely

to calculate the accurate

dΔ .

5 CONCLUSIONS

In this paper the influence of time gap on user’s

typing performance is studied, and a unique Time

Gap model is developed. Experimental results show

that the current keystroke’s predecessor affected the

user’s typing behaviour, and the Time Gap between

two consecutive keystrokes is influenced by current

symbol’s predecessor. Inspired by this conclusion, a

fundamental PTG model is developed. Its

experimental results indicate that the correction rates

predominantly lie in between 65% and 90% with the

current testing sample.

Furthermore, an innovative Distance, Angle and

Time Gap PNN model based on Probabilistic Neural

Network is developed to simulate a specific user

typing behaviour – ‘Hitting Adjacent Key Errors’

based on unique factors such as key distances.

Results demonstrate that about 70% of all tests score

above Basic Correction Rate. Results also show a

very unstable trend of user’s ‘Hitting Adjacent Key

Errors’ behaviour, which suggest that several

training trials with a random dataset selection

strategy could be applied.

PATTERNS IDENTIFICATION FOR HITTING ADJACENT KEY ERRORS CORRECTION USING NEURAL

NETWORK MODELS

ACKNOWLEDGEMENTS

The research is funded by Disability Essex and

Technology Strategy Board. Thanks to Richard

Boyd, Pete Collings and Ray Mckee for helpful

advice and discussions.

REFERENCES

Artificial neural network, [online], 31 Dec. 2010,

available: http://en.wikipedia.org/wiki/Artificial_

neural_network [12 January 2010]

David J. Ward, Alan F. Blackwell at el. (2000). ‘Dasher-a

Data Entry Interface Using Continuous Gestures and

Language Models’, UIST '00 Proceedings of the 13th

annual ACM symposium on User interface software

and technology

Disability Essex, http://www.disabilityessex.org [accessed

18 January 2009]

Karim Ouazzane, Jun Li and Marielle Brouwer (2008). ‘A

hybrid framework towards the solution for people with

disability effectively using computer keyboard’,

IADIS International Conference Intelligent Systems

and Agents 2008, pp. 209-212

Knowledge Transfer Partnership, http://www.

ktponline.org.uk/ [accessed 18 January 2009]

Paul M. Fitts (1954). ‘The information capacity of the

human motor system in controlling the amplitude of

movement’, Journal of Experimental Psychology,

volume 47, number 6, June 1954, pp. 381-391

Prototype, [online], n.d., available: http://www.sensory

software.com/prototype.html [accessed 03 March

2008]

R. W. Soukoreff, & I. S. MacKenzie, n.d. KeyCapture

[online], available: http://dynamicnetservices.com/

~will/academic/textinput/keycapture/ [accessed 18

January 2009]

R. W. Soukoreff, & I. S. MacKenzie (2003). ‘Input-based

language modelling in the design of high performance

text input techniques’, Proceedings of Graphics

Interface 2003, 89-96.

The Dasher Project, [online], 14 Nov. 2007, Inference

Group of Cambridge, available: http://www.inference.

phy.cam.ac.uk/dasher/ [accessed 03 March 2008]

Unary coding, [online], 23 December 2009, available:

http://en.wikipedia.org/wiki/Unary_coding [12

January 2010]

Virtual key codes [online], available:

http://api.farmanager.com/en/winapi/virtualkeycodes.h

tml [accessed 05 February 2009]

ICEIS 2011 - 13th International Conference on Enterprise Information Systems