Table of Contents 1. Introduction
1
1.1 purpose
1
1.2 Scope
2
1.3Definations , Abbrevations
2
1.4 Refrences
3
1.5 Overall
4
2. The Overall Description
5
2.1 Product Perspective
5
2.2 Product Functions
5
2.3 User Characteristics
6
2.4 General ConstraintsAssumptions and Dependcies
6
3. Particular Requirments 3.1 External Interface Requirments
6 7
3.1.1 Sytem Interfaces
7
3.1.2 User Interfaces
7
3.1.3 Hardware Interfaces
7
3.1.4 Software Interfaces
8
3.1.5 Communications Interfaces
8
3.2 Functional Requirements
8
3.2.1
8
3.2.2
10
3.3 Use Cases 3.3.1 Use Case #1
3.4 Classes / Objects 3.4.1
10 10
11 11
Handwritten Digit Recognition based on Multifeatures 3.5 Non-Functional Requirements
12
3.5.1 Performance
12
3.5.2 Reliability
13
3.5.3 Availability
13
3.5.4 Security
13
3.5.5 Maintainability
13
3.5.6 Portability
13
3.6 Logical Database Requirements
14
3.7 Design Constraints
14
3.7.1 Standards Compliance
14
4. Analysis Models
15
4.1 Sequence Diagrams
16
4.2 Activity Diagram
17
4.3 Data Flow diagram(dfd)
18
4.4 State Transition diagram(std)
18
5. Supporting Information
18
Appendix A – A – Background Background Research on: Appendix B – Data Dictionary
18 22
Appendix C – C – Porposal Porposal
22
ii
Handwritten Digit Recognition based on Multifeatures
1. Introduction
In the running scene, there is developing interest for the product frameworks to perceive characters and digit in PC framework when data is looked over paper records as we realize that we have number of daily papers and books which are in printed design identified with various subjects. Nowadays there is a colossal interest in "putting away the data accessible in these paper reports into a PC stockpiling plate and after that later reusing this data via seeking prepare likewise an advantage to perceive that from which catchphrase it is related". One basic approach to store data in these paper records into PC framework is to first output the reports and afterward store them as IMAGES. Be that as it may, to reuse this data it is extremely hard to peruse the individual substance and seeking the substance shape these reports line-byline and word-by-word. The explanation behind this trouble is the text style attributes of the characters in paper archives are distinctive to textual style of the characters in PC framework, Is there any framework for manually written watchwords?. Thus, PC can't perceive the characters and digit while understanding them. This idea of putting away the substance of paper archives in PC stockpiling spot and after that perusing and seeking the substance of transcribed is called DOCUMENT PROCESSING. Here and there in this report handling we have to prepare the data that is identified with dialects other than the English on the planet. For this record preparing we require a product framework called CHARACTER AND DIGIT RECOGNITION SYSTEM. This procedure is additionally called DOCUMENT IMAGE ANALYSIS (DIA).
1.1 Purpose
This record is the product Requirements Specifications for HWR 1.0. This portrays the capacities and execution prerequisites of the product about which we make the item. This task makes utilization of effective picture preparing systems and neural systems (outspread Basis and BackPropagation).Handwriting acknowledgment (or SRS Document 1.0
Page 1
Handwritten Digit Recognition based on Multifeatures HWR) is the capacity of a PC to get and decipher clear manually written information from sources, for example, paper documents, photographs, touch-screens and different gadgets, and recognize the digit on the off chance that it is, and after that change it in the appropriate example about which it is understandable.
1.
The essential goal is to accelerate the procedure of character and digit acknowledgment in archive preparing. As it is multi framework programming so therefore the framework can prepare colossal number of records with-in less time and subsequently spares the time.
2.
Since our character acknowledgment depends on a framework base, it expects to perceive various heterogeneous characters and digit that have a place with various widespread dialects with various textual style properties and arrangements. It likewise perceived the spell checking of characters.
1.2 Scope
The extent of our item Handwritten Digit and Character Recognition and multifeatures is to give a productive and upgraded programming apparatus for character investigation and for the clients to perform Document Image Analysis, report handling by perusing and perceiving the characters in examination, scholastic, administrative and business associations that are having substantial pool of archived, filtered pictures. Regardless of the extent of records and the kind of characters in archives, the item is remembering them, seeking them and handling them quicker as per the requirements of nature which is the need of nowadays. 1.3 Definitions, Acronyms, and Abbreviations.
Neural Networks: An Artificial Neural Network (ANN) is an information taking care of perspective that is moved byorganicsensory frameworks, for instance, the cerebrum, process information.According to taking a gander at nervious.1.5 Overview: SRS Document 1.0
Page 2
Handwritten Digit Recognition based on Multifeatures 1.4 References:
1. Recognition of Handwritten text
using Proximal Support Vector Machine
bySwapnaPrava Ekka (2014) 2. Learning algorithm for Classification: A comparison on hand written digits and character recognitionbyYannLecun 3. Representation and Recognition of Handwritten Digits and Character Using DeformableTemplates (1997)Anil K. Jain, Fellow, IEEE, and Douglas Zongker
4. Using generative models for hand written digit recognitation Michael Revow, christopherK.I. Williams and geoffrayE.Hinton (1996) 5. Handwritten Digits RecognitionbyGaurav Jain, Jason Ko (2008) 6. Recognizing Handwritten Digits and Characters byVishnuSundaresanJasper Lin (2015) 7. HANDWRITTEN DIGIT CLASSIFICATION byAndreaGiuliodori, Rosa Lillo and Daniel Peña (2011)
1.4Overview: Till now, we have given brief prologue to HWRC 1.0. The point of the record is to give a few subtle elements into framework, and a few imperatives of it. Our way to deal with take care of this issue of manually written numeral and Character acknowledgment can be extensively partitioned into three pieces: 1. Digitization/Pre-processing 2. Feature Extraction utilizing PCA 3. Classification utilizing 1‐ Nearest Neighbor calculation The piece chart for the framework is demonstrated as follows
SRS Document 1.0
Page 3
Handwritten Digit Recognition based on Multifeatures
3. The Overall Description:
Principally the manually written digit acknowledgment framework can Acquaintance the digits (0-9) which are composed by the people or machine in non-writing technique. .The assignment of manually written digit acknowledgment, utilizing a classifier, has extraordinary significance and utilize, for example, – internet penmanship acknowledgment on PC tablets, perceive postal divisions of sorting mail postal, preparing banks checks sums, numerics passages in structures topped off byhand etc Additionally individuals compose the same digit in various ways ‐ the digit "1" is composed as '1', '1', "1" or '1'. Essentially 7 might be composed . At long last the uniqueness and assortment in the penmanship of various people additionally impacts the arrangement and appearance of the digits. The framework just acknowledges the digits (0-9) if the information is something else it creates mistake for depiction. The framework depicts the edges precisely proposed the digit's limit so it is regularly utilized for HDR.
2.1 Product Perspective
HWR is first of its kind and doubtlessly first open source endeavor to perceive the essayist in view of his composition as every individual has an alternate written by hand style. The point SRS Document 1.0
Page 4
Handwritten Digit Recognition based on Multifeatures of the venture would to add to a framework which can be utilized effectively everywhere associations, for example, banks and even in legal divisi on.
2.2 Product Functions
The real capacity of HWRC is to perceive an essayist from the pool of author. Individual confirmation is essential at numerous associations; this venture might be utilized as a part of such places. Individual recognizable proof is likewise imperati ve in measurable division.
2.3 User Characteristics
1. Large Organizations:
Once totally built up, the framework can be utilized as a part of legal office, banks and wrongdoing scene division.
2. Academic Organizations:
College and Students are relied upon to add to an extensive class of clients of this framework to get the comprehension of significant parts utilized as a part of the system like picture handling strategies, neural systems and manually written notes.
SRS Document 1.0
Page 5
Handwritten Digit Recognition based on Multifeatures
3General Constraints, Assumptions and Dependencies
The framework must be prepared completely before use and clients knows the framework particular. When framework is prepared, before leaving you are required to spare the framework, so the framework stacked for further use. Also has the upgraded quality.
3. Particular Requirements
The best and key bit to any SRS. These client necessities will be exemplified inside of Section 2, yet this domain will give the D-prerequisites that are utilized to organize the endeavor of the thing use, setup and testing. Every key in this area ought to be: •Correctness •Tracable •Unabmigious •Verfiy •Complete •Consistent (Having Integrity) •Uniquely identifiable (as a general rule through numbering like 3.4.5.6) Thought should be paid to the intentionally organize the necessities displayed around there with the objective that they may viably got to and got on. In addition, this SRS is not the item arrange document, hence one should avoid the slant to over-oblige (and along these lines diagram) the item wander within SRS.
3.1 External Interface Requirements 3.1.1 System Interface
Generally there would be one in control who working the structure and tell whether a particular individual is affirmed or not. Not most of the customers need to know not the system as showed by all specific.
3.1.2 User Interfaces
Client Interface:
SRS Document 1.0
Page 6
Handwritten Digit Recognition based on Multifeatures
3.1.3 Hardware Interfaces:
A decent scanner is required so that the composed specimens can be examined and given as info the framework. On the off chance that scanner is not accessible then just composed characters or digits are perused by framework with mouse or advanced pen.
3.1.4 Software Interfaces:
Some picture preparing programming, for example, gimp might be utilized to change over the pictures which are not bmp configuration to bmp pictures so they get to be comprehensible by the project.
3.2 Functional Requirements:
This area depicts particular elements of the DCRS programming venture. On the off chance that fancied, a few prerequisites might be indicated in the utilization case organize and recorded in the Use Cases diagrams.
3.2.1 Functional Requirement or Feature
SRS Document 1.0
Page 7
Handwritten Digit Recognition based on Multifeatures 3.2.1.1 Introduction:
The point of this venture is to actualize a grouping calculation to perceive manually written character and digit It has been appeared of example acknowledgment performs the best for all grouping issues reliably. Subsequently, the extent of the undertaking likewise incorporated the rudimentary study the distinctive classifiers and blend techniques, and assess the admonitions around their execution in this specific issue of manually written digit acknowledgment. We could accomplish a precision rate of 78.4%.
3.1.1.2 Inputs Digit, order, pictures
General Constraints, Assumptions and Dependencies The framework must be prepared completely before use and clients knows the framework particular. When framework is prepared, before leaving you are required to spare the framework, so the framework stacked for further use. Also has the upgraded quality.
SRS Document 1.0
Page 8
Handwritten Digit Recognition based on Multifeatures 3.1.1.3Processing
3.1.1.2 Outputs
SRS Document 1.0
Page 9
Handwritten Digit Recognition based on Multifeatures
3.2.1.5 Error Handling
Expectation and Try Catch is used to control Error Handling, In character input if user enter wrong spellings the system catch it and show error to check spelling mistakes. 3.3 Use Case 3.3.1 Use Case
Operations of Digit and Characters recognition
SRS Document 1.0
Page 10
Handwritten Digit Recognition based on Multifeatures 3.4 Class /Object:
3.4.1 Class/Object
An arrangement of all points or classification of things having some property or trait in like manner and separated from others by kind, sort, or quality. 3.4.1.1 Class Character Attributes
1. A to Z 2. a to z Function
Character Processing Differentiate from the digits
3.4.1.2 Class Digits Attribute
1
0 to 9 digit
Function
Digit Processing Differentiate from the Character 4.1.1.3 Class
Spellchecker
Function
Character Spell Processing
3.5 Non-Functional
Non-utilitarian essentials may for eixit running with qualities. A significant part of the time the essentials must master at a structure level of wide rather than at a unit level. State the essentials in the running with locales in quantifiable terms..
SRS Document 1.0
Page 11
Handwritten Digit Recognition based on Multifeatures 3.5.1 Performance
Execution relying upon the connection, high PC execution might include one or a greater amount of the accompanying: Short reaction time for a given bit of work. Execution is described by the measure of valuable work achieved by a PC framework or PC system contrasted with the time and assets utilized 3.5.2 Reliability
Unwavering quality is a property of any PC related part (programming, or equipment, or a system, for instance) that reliably performs as indicated by its determinations. It has for some time been viewed as one of three related qualities that should be considered when making, purchasing, or utilizing a PC item or part. 3.5.3 Availability
In PC frameworks and systems administration, accessibility is a general term that is utilized to depict the measure of time over a one-year period that the framework assets is accessible in the wake of part disappointments in the system. A framework constantly accessible with its all assets are viewed as fruitful. 3.5.4 Security
In registering, security (or PC security) is the strategies for guaranteeing that information put away in a PC can't be perused or bargained by any people without approval. Most PC efforts to establish safety include information encryption and passwords. Information encryption is the interpretation of information into a structure that is indiscernible without a disentangling system. A watchword is a mystery word or expression that gives a client access to a specific project or framework.
3.5.5 Maintainability
Viability is characterized as the likelihood of performing an effective repair activity inside of a given time. As such, practicality measures the straightforwardness and pace with which a framework can be restored to operational status after a disappointment happens.
3.5.6 Portability
SRS Document 1.0
Page 12
Handwritten Digit Recognition based on Multifeatures Convenience is a trademark credited to a PC program in the event that it can be utilized as a part of working frameworks other than the one in which it was made without requiring major revamp. Porting is the assignment of doing any work important to make the PC program keep running in the new environment. 3.6 Logical Database Requirements:
•Types of utilized data Utilization of frequency Abilities Data elements & connections Limitations of integrity Data maintenance necessities Protection of security dangers
3.7 Design Constraints:
Determine outline imperatives that can be forced by different models, equipment confinements, and so on. 3.8 Standards Compliance:
Decide the essentials got from existing pages or regulation . They may include: Report bunch Naming of data Procedures of accounting Tracing of audit Case in point, this could demonstrate the essential for programming to take after get ready activity. Such takes after are required for a couple of uses to meet minimum regulatory or money related standards. A survey take after essential may, for case, express that all
SRS Document 1.0
Page 13
Handwritten Digit Recognition based on Multifeatures movements to a money database must be recorded in a take after report with already, then sometime later values. 4. Investigation Models
His framework is for pictures and manually written digits and characters to perceived and check the spelling of characters composed by hand so that is the reason it is called multi highlights programming.
SRS Document 1.0
Page 14
Handwritten Digit Recognition based on Multifeatures 4.1Use Case Diagram:
SRS Document 1.0
Page 15
Handwritten Digit Recognition based on Multifeatures 4.2 State Diagram
4.3 Activity Diagram
SRS Document 1.0
Page 16
Handwritten Digit Recognition based on Multifeatures
4.4 Data Flow Diagrams (DFD)
SRS Document 1.0
Page 17
Handwritten Digit Recognition based on Multifeatures
5. Supporting Information
Appendix A – Background Research on: Not avail Appendix B – Data dictionary:
A:
Abstract Availability Acronyms Abbreviations Academic Organizations: Assumptions Analysis Models
B:
Background
C:
Class / Object: D:
Definitions Dependencies Data Flow Diagrams (DFD) Design Constraints
SRS Document 1.0
Page 18
Handwritten Digit Recognition based on Multifeatures E:
External Interface Requirements Error Handling F:
Functional Requirements Functional Requirement or Feature Feature Extraction using Principal Component Analysis:
G:
General Constraints,
H:
Hardware Interfaces: Hand written digit and characters recognition and spell checker (HWDCRS)
I:
Introduction Interfaces Introduction Inputs
K:
Keywords:
L:
Logical Database Requirements
M:
Motivation Maintainability METHADOLOGY:
SRS Document 1.0
Page 19
Handwritten Digit Recognition based on Multifeatures N:
Neural Networks Non-Functional Requirements
O:
Overview: Overall Description P:
Purpose Product Perspective Product Functions Processing Performance Portability Problem statement Preprocessing R:
References: Reliability Research question Research objectives
S:
Specific Requirements System Interface Security SRS Document 1.0
Page 20
Handwritten Digit Recognition based on Multifeatures SCOPE: Straight lines detection Sequence Diagrams: State-Transition Diagrams (STD) Supporting Information System Overview Standards Compliance
U:
User Characteristics Use Cases
Appendix C – Proposal
Handwritten Text Recognition Based On Multifeatures Abstract
Design acknowledgment is one of the real difficulties in insights structure. Its objective is the component extraction to order the examples into classes. An understood illustration in this field is the transcribed digit acknowledgment where digits must be allocted into one of the 10 classes utilizing some grouping technique. Our motivation is to present option grouping strategies in light of measurable methods. We demonstrate a correlation between a multivariate and a probabilistic methodology, reasoning that both techniques give comparative results as far as test-blunder rate. Examinations are SRS Document 1.0
Page 21
Handwritten Digit Recognition based on Multifeatures performed on the known MNIST and USPS databases in double level picture. At that point, as an extra commitment we acquaint a novel strategy with pairs pictures, taking into account measurable ideas related to the composed hint of the digit. Key words
Digit, arrangement, pictures Introduction
In the running scene, there is developing interest for the product frameworks to perceive characters in PC framework when data is looked over paper archives as we realize that we have number of daily papers and books which are in printed design identified with various subjects. . The stored data in these paper archives into PC framework is to first sweep the reports and after that store them as IMAGES. Be that as it may, to reuse this data it is exceptionally hard to peruse the individual substance and looking the substance shape these reports line-by-line and word-by-word. The purpose behind this trouble is the textual style qualities of the characters in paper records are diverse to textual style of the characters in PC framework. Subsequently, PC can't perceive the characters while understanding them. This idea of putting away the substance of paper archives in PC stockpiling spot and afterward perusing and seeking the substance is called DOCUMENT PROCESSING. Some of the time in this report preparing we have to handle the data that is identified with dialects other than the English on the planet. For this archive preparing we require a product framework called CHARACTER RECOGNITION SYSTEM and SPELL CHECKER. This procedure is likewise called DOCUMENT IMAGE ANALYSIS (DIA).
Foundation and Motivation
Handwritten composing acknowledgment
of character has been produce 1980,s..The
assignment is written by handdigit and .Character acknowledgment, utilizing a classifier, has extraordinary significance and utilize, for example, – web penmanship acknowledgment on PC tablets, perceive postal divisions sorting on mail postal.., handling checks bank sums, didit passages in structures topped off hand (for instance ‐ tax documents) et cetera. There are distinctive difficulties confronted while endeavoring to take care of this issue. The digits and character is not generally of size of same thickness or introduction and position in respect of edges. Our is objective of actualize an example characterization technique to perceive the transcribed digitsAnd .
SRS Document 1.0
Page 22
Handwritten Digit Recognition based on Multifeatures Character gave in the MINIST information set of pictures of manually written digits (0 ‐9). The information set utilized for our application is made out of 300 preparing pictures and 300 testing pictures, and is a subset of the MNIST information set [1] (initially made out of 60,000 preparing pictures and 10,000 testing pictures). Every picture is a 28 x 28 grayscale (0‐255) named representation of an individual digit.
Process Overview :
Our way to deal with take care of this issue of written by hand numeral acknowledgment can be extensively isolated into three squares: Pre‐Processing/Digitization Highlight Extraction utilizing PCA Grouping utilizing 1 ‐ Nearest Neighbor calculation PreProcessing
Amid our underlying audit of the different classifier techniques, we attempted the methodology of preparing the preparation set pictures to decrease the information by thresholding the offered picture to a twofold picture. Fig.2 speaks to the specimen pictures taken from the MNIST database [1].
Amid our underlying audit of the different classifier strategies, we attempted the methodology of preparing the preparation set pictures to lessen the information by thresholding the offered picture to a twofold picture. Fig.2
SRS Document 1.0
Page 23
Handwritten Digit Recognition based on Multifeatures
Fig 2 We additionally took a gander at different preparing strategies, for example, edge‐detection, and diminishing of the digit picture [2] to get a skeleton of the digit. This methodology of gaining the skeleton of the digit is broadly utilized as a part of the classifiers which fundamentally answer upon a well ‐defined info picture for their exactness. PCA is the all encompassing methodology that concentrates eigendigits taking into account the general .
Highlight Extraction utilizing Principal Component Analysis :
Vital section examination (PCA) is a significant multivariate data examination methodology which is experienced into a combination of zones in neural frameworks, signal taking care of, and machine learning. It is an unsupervised system for decreasing the dimensionality of the present data set and evacuating key information. PCA does not use any yield information; the standard to be expanded is the distinction. PCA can be associated with monetarily address the data digit pictures and Character by expecting them onto a low ‐dimensional space constituted by somewhat number of reason pictures [3]. These reason pictures or the "eigendigits" are controlled by finding the most colossal eigenvectors of the pixel sagacious covariance system, after mean centering the data for each property. After projection, we use the 1 ‐ NN classifier to aggregate the digit in the low dimensional space. PCA diminish the estimations of the dataset from 784 to a lower worth – for straightforwardness of figuring. Straight lines identification .
SRS Document 1.0
Page 24
Handwritten Digit Recognition based on Multifeatures The principle precept of the HT is to identify the event of focuses (pixels for us) in a picture, lying on a straight line. The mathematical statement for a straight line is spoken to in the Cartesian organizes as (4) yi = a xi + b; what's more, is graphically plotted for a couple of focuses (x1; y1) and (x2; y2) .. The HT plans to and focuses with directions (x; y) that fulfill the mathematical statement 4. There exist infantine lines which go through a specific point (xi; yi) in the Cartesian plane, however stand out line satiates the mathematical statement for specific estimations of parameters an and b. Also, focuses lying on the same straight line in Cartesian plane can be spoken to in the space of parameters an and b as it is appeared in Figure 4b. That implies, two focuses lying on the same straight line with parameters an and (a) Cartesian plane (b) Parameter space The estimation of HT, the parameters and can be limited to and lines With a specific slant or position in a picture. Additionally the base number of pixels required to adjust a line (the base estimation of a top p) can be resolved.
In the wake of making a few examination in our work, we consider fascinating to and vertical(90o), flat (0o) and corner to corner (45o) lines to differentiate digits. Since pictures have little size we select lines with no less than two pixels. Accordingly, the Hough change recognized every single conceivable line with those attributes in four stages. (1) In the first stage the parallel picture is part on a level plane into two rectangular equivalent amounts of and the biggest level line is enlisted from every part of the picture. (2) In the second phase , the paired of picture is vertically into two rectangular a balance of and the biggest vertical line is distinguished from every part of the picture. (3) In the third stage, the picture is isolated by its main corner to corner and the largest upper and lower parallel to this slanting are found. (4) In the four stage, the picture is isolated by its optional inclining and the largest upper and lower parallel to this slanting are found the highlights considered in the classification were gotten from the information of the chose lines. Each line has two focuses, the begin and the end-point. The coordinates (xi; yi) which indicate the begin point and end-purpose of a straight line are what we call straight (S) to determine the direction of lines of 0o and 90o. In addition, we called askew (D) to allude the direct ions of a slanting lines (with 45o). We SRS Document 1.0
Page 25
Handwritten Digit Recognition based on Multifeatures likewise incorporate the length (Le) of every line as an extra variable. Along these lines, we have an aggregate of 5 variables for each line . Since we utilize four straight lines and four diagonals todined the state of a picture, we record 40 coming about qualities for every picture.
(a)
Cartesian
plane
Problem Statement:
Penmanship number acknowledgment is a testing issue specialists had been exploration into this region for so long particularly in the late years. In our study there are numerous fields worry with numbers, for instance, checks in banks or perceiving numbers in auto plates, the subject of digit acknowledgment shows up. A framework for perceiving detached digits might be as a methodology for managing such application. At the end of the day, to give the PC a chance to comprehend the handwritten digit that is composed physically by clients and perspectives them as per the system control.
RESEARCH OBJECTIVES:
The primary goal for our topic Handwritten digits. For instance, diverse clients had their own particular penmanship styles where here the SRS Document 1.0
Page 26
Handwritten Digit Recognition based on Multifeatures fundamental test tumbles to give PC framework a chance to comprehend these diverse penmanship styles and remember them as standard written work.
REASERCH QUESTIONS:
The principle issues is perceive the digits and separate between digits.
METHADOLOGY:
We exhibited a framework for managing such issue. The framework began by obtaining a picture containing digits, this picture was digitized utilizing some optical gadgets and in the wake of applying a few improvements and alterations to the digits inside of the picture
SCOPE:
We added to a framework for written by hand digits and character acknowledgment. What's more, we productively pick a strategy to fit our requests. Our framework effectively outlines and execute a neural system which proficiently abandon requests, after that the framework can comprehend the transcribed digits that was composed physically by clients. This venture is useful to perceive written by hand digits by client.
REFERENCES :
1. Catalin I. Tomai, Bin Zhang and Sargur N. Srihari “Discriminatory Power of Handwritten Words for Writer Recognition” Proceedings of the 17th International Conference on Pattern Recognition (ICPR’04) 2. Marius Bulacu, and Lambert Schomaker “Text-Independent Writer Identification and Verification Using Textural and Allographic Features” IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 29, NO. 4, APRIL 2007 3. R. Plamondon and S.N. Srihari, “Online and Offline Handwriting Recognition: A Comprehensive Survey,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22, no. 1, Jan. 2000. 4. NEURAL
NETWORK
MODELS
AND
THEIR
APPLICATION
TO
HANDWRITTEN DIGIT RECOGNITION by Thaddeus F. Pawlicki Dar-Shyang Lee 5. Jonathan J. Hull Sargur N. Srihari Department of Computer Sciences
SRS Document 1.0
Page 27
