voice biometric report

7/28/2019 Voice Biometric Report

1/86

Voice BiometricSecured Control Unit

ABSTRACT

The proposed Voice Biometric Security for Industrial control verifies a

persons claimed identity on the basis of his/her speech phrases. It is word-

dependent, which requires the speaker to say key words or sentences having the

same words for both enrolling and verifying trials.

This system uses template-matching approach to authenticate user. The

system processes voice signal of the user, which is given as input and prepares

a template by using Mel Frequency Cepstrum Coefficients (MFCC) technique

and verifies it with the already existing template by using Vector Quantization

(VQ) technique. Then it declares the person as valid person or an imposter

depending on the template matching.

Once Verification is achieved it allows the user get into a control panel and

control LCD device from PC.

Department Of Electronics & CommunicationAppa Institute of Engineering and Technology,

Gulbarga

1


2/86


CHAPTER 1

Introduction

1.1 General introduction

The Voice Biometric Security for Industrial control is an application of

Biometrics. Biometrics refers to the automatic identification of a person based on

his/her physiological or behavioral characteristics.

Much research has been done in the area of speaker verification using

cepstral analysis, end point detection algorithms, pattern recognition, neural

networks, stochastic models and many-distance measuring algorithms. However

as per media reports, the satisfactory performance of the existing systems is still

a matter of great concern because of the considerable number of false

acceptances and false rejections.

Speaker verification is a difficult task and it is still an active research

area. A speech technology research center at Sydney University, Australia is

actively involved in speech recognition, human speech perception, and natural

language processing. Microsoft is also an active participant of speech technology

research whose aim is to produce a complete speech enabled computer. To

reach this stage, Microsoft is also contributed its research and development in


Gulbarga

2


3/86


several other areas including noise robustness, automatic grammar learning and

language modeling.

The Speaker Verification can be used in many areas. Some of them are

Access control to computers, cellular phones, databases, Access control to

professional or wide public sites, Protection of confidential data, Remote access

to computer networks, Electronic commerce, Forensic, Automatic door opening

systems when person arrives, Telephone-banking transactions and Voice

commands on cellular phones. Many organizations like banks, institutions,

industries etc are currently using this technology for providing greater security to

their vast databases.

Given the voice input, the objective of the proposed word-dependent

Voice Biometric Security for Industrial control is to verify whether the speaker is

who he/she claims to be. The system processes voice signal of the user, which is

given as input and finds the Mel Frequency Cepstrum Coefficients. Then it

generates a template, which is called a codebook, an array of acoustic vectors.

In the enrollment phase the user codebook is saved in the system. In the

verification phase the user codebook is compared against the claimed users

actual codebook, which is stored in the system during the enrollment phase. If

the difference is below the threshold value the user is authenticated else the user

is not authenticated.


Gulbarga

3


4/86


1.2 Statement of the problem

With the increased use of computers as vehicles of information

technology, it is necessary to restrict unauthorized access to the systems or to

sensitive/personal data.

The traditional methods of authentication involving passwords, PIN

numbers and tokens have some drawbacks. Because PINs and passwords may

be forgotten, and token-based methods of verification like passports and driver's

licenses may be forged, stolen, or lost. Therefore, there is a need for a better

method of authentication like voice verification which can eliminate these

drawbacks, so that, the person to be verified is required to be physically present

at the point-of-verification.

The other authentication mechanisms (such as face recognition, iris

recognition etc.) need a lot of equipment and conditions to work properly. So,

there is a need for the cheaper authentication system like speaker verification,

which works with less equipment.


Gulbarga

4


5/86


1.3 Objective of the study

The objective of the study is to build a Voice Biometric Security for

Industrial control for the following reasons:

Given a sample of somebody saying a particular piece of text/passphrase (for example, the person's name), the system should be able to

verify whether the speaker is who he/she claims to be.

Furthermore, the system should be able to catch ``imposters'' who tryto say somebody else's pass phrase.

Once the person is authenticated the system allows him to control anydevice such as LCD, STEPPER MOTOR etc.


Gulbarga

5


6/86


1.4 Scope of the study

As with all new security solutions, the application of voice biometrics

technology within a given corporate environment will be based on set policies

and known sensitivities. To aid its acceptance, voice verification can be

combined with more traditional security features to provide an additional layer of

authentication. The commercialization of voice technology offers network

administrators, the new opportunities to enhance advanced user authentication

methods, password control and innumerable user identification and network

security applications. In a society where telecommunications and electronic

commerce are the norm, the need to protect sensitive information will

undoubtedly act as a catalyst to greater use of biometrics and in turn, result in

improved technology.

The present system can be extended to become an integral part of speech

interfaces with voice identification. This is a complex task in which an unknown


Gulbarga

6


7/86


voice template has to be compared with a huge set of voice templates to make it

a voice verification product. This product can be used as part of multi-model

verification, which includes face, fingerprints and iris scan etc.

In this way, voice biometrics system is expected to create new services

that will make our daily lives more convenient.

1.5 Review of literature

As per the needs and requirement for the study and implementation of Voice

Biometric Security for Industrial control, information is gathered from various possible

sources. In fact, there is no single literature that could serve as a complete idea of

speaker recognition.

Before actually implementing the Voice Biometric Security for Industrial control, I

needed to study about the recent progress, current applications and future trends in the

area of speaker recognition. I have gone through [4], to know what is currently needed in

the industry market.

Since Voice Biometric Security for Industrial control is an application of a

kind of Biometric method of identification which is used to authenticate a person

on the basis of his/her behavioral characteristic like voice, I needed [9] and [10]

to get a range of information about biometrics and speech technology.


Gulbarga

7


8/86


The information needed for the methodology, which involves Mel

Frequency Cepstrum Coefficients technique for Feature extraction process and

Vector Quantization technique for Feature matching process, is taken from [5]

and [6]. Feature extraction is the process that extracts a small amount of data

from the voice signal that can later be used to represent the speaker. Feature

matching involves the actual procedure to verify the speaker by comparing

extracted features from his/her voice input with the claimed one which is stored in

the database.

To understand how to proceed in steps for developing the software for the

system, and to analyze and design the system in object-oriented paradigm, [1] and

[2] books have been referred. The Use-case driven object-oriented analysis and

object-oriented design are done in a modeling language called Unified Modeling

Language (UML). And the book [3] was found very helpful to implement the whole

idea in Visual C++.

The rest of the information needed about speaker recognition during

developing the system, including its benefits and natural advantages are studied

from [7], and [8].


Gulbarga

8


9/86


1.6 Methodology

1.6.1 Organization of the system

The Voice Biometric Security for Industrial control mainly consists of two

phases: Enrollment phase and the Verification phase. The organization of the

system is shown below diagrammatically through the figures Fig.1.6.1 (a) and

Fig.1.6.1 (b).

During the first phase, speaker enrollment as shown in the Fig.1.6.1 (a),

features are extracted from the input speech signal given by the speaker by a

process called Feature extraction, and are modeled as a template. The modeling

is a process of enrolling speaker to the verification system by constructing a

model of his/her voice, based on the features extracted from his/her speech

sample. The collection of all such enrolled models is called speaker database.


Gulbarga

9


10/86


Fig.1.6.1(a): Schematic flow of Enrollment Phase

In the second phase, verification phase as shown in the Fig.1.6.1 (b), features

are extracted from the speech signal of a speaker and these current features are

compared with the claimed features stored in the database by a process called

Feature matching. Based on this comparison the final decision is made about the

speaker identity.


Gulbarga

10

Decision


11/86


Fig.1.6.1 (b): Schematic flow of Verification Phase

Both these phases include Feature extraction, which is used to extract

speaker dependent characteristics from speech. The main purpose of this

process is to reduce the amount of test data while retaining speaker

discriminative information.

1.6.2 Feature extraction process

1.6.2.1 Introduction

In speaker recognition first we convert the speech waveform to some type

of parametric representation (at a considerably lower information rate) for further

analysis and processing. This is often referred as the signal-processing front

end.

A wide range of possibilities exist for parametrically representing the

speech signal for the speaker recognition task, such as Linear Prediction Coding

(LPC), Mel Frequency Cepstrum Coefficients (MFCC), and others. MFCC is

perhaps the best known and most popular, and this will be used in this project.

MFCCs are based on the known variation of the human ears critical

bandwidths with frequency; filters spaced linearly at low frequencies and


Gulbarga

11


12/86


logarithmically at high frequencies have been used to capture the phonetically

important characteristics of speech. This is expressed in the Mel-frequency

scale, which is linear frequency spacing below 1000 Hz and a logarithmic

spacing above 1000 Hz. The process of computing MFCCs is described in more

detail next.

1.6.2.2 Mel-frequency cepstrum coefficients

processor

A block diagram of the structure of an MFCC processor is given in

Fig.1.6.2.2. The speech input is recorded at a sampling rate of 16 KHz. This

sampling frequency was chosen to minimize the effects of aliasing in the analog-

to-digital conversion.


Gulbarga

12

framecontinuous

speechFrame

Blocking FFT

spectrum

Mel-frequencyWrapping

Cepstrum

melmel

spectrumcepstrum

Windowing


13/86


Fig.1.6.2.2: Block diagram of the MFCC processor

The processor takes continuous speech signal as input and generates

mel-frequency cepstrum coefficients as outputs. It uses the following 5 steps to

accomplish this task:

(1)Frame blocking:

In this step the continuous speech signal is blocked into frames of N=16

samples, with adjacent frames being separated by M=10. The first frame

consists of the first N samples. The second frame begins M samples after the

first frame, and overlaps it by N - M samples. Similarly, the third frame begins

2M samples after the first frame (or M samples after the second frame) and

overlaps it by N - 2M samples. This process continues until all the speech is

accounted for within one or more frames.

(2)Windowing:

The next step in the processing is to window each individual frame so as

to minimize the signal discontinuities at the beginning and end of each frame.

The concept here is to minimize the spectral distortion by using the window to

taper the signal to zero at the beginning and end of each frame. If we define the

window as 10),( Nnnw , where N =16 is the number of samples in each

frame, then the result of windowing is the signal:


Gulbarga

13


14/86


10),()()( = Nnnwnxny ll

(1.1)

I have used a Hamming window here, which has the form:

10,1

2cos46.054.0)(

= Nn

N

nnw

(1.2)

(3)Fast Fourier Transform (FFT):

The next processing step is the Fast Fourier Transform, which converts

each frame of N =16 samples from the time domain into the frequency domain.

The FFT is a fast algorithm to implement the Discrete Fourier Transform (DFT)

which is defined on the set of N samples {xn}, as follow:

=

==1

0

/21,...,2,1,0,

N

k

Njkn

kn NnexX

(1.3)

Note that we use j here to denote the imaginary unit, i.e. 1=j . In

general Xns are complex numbers. The resulting sequence {Xn} is interpreted as

follow: the zero frequency corresponds to n = 0, positive frequencies 2/0 sFf


15/86


The FFT algorithm I have used is discovered by J.W. Cooley and J.W.

Tukey in 1965. This algorithm is based on the following three steps:

1. Decompose an N point time domain signal into N signals each containing a

single point.

An Interlaced decomposition is used each time a signal is broken in two, that is,

the signal is separated into its even and odd numbered samples. There are Log2 N

stages required in this decomposition. Since I have used a 16 point signal, it

requires 4 stages.

2. Find the spectrum of each of the N point signals.

Nothing is required to do this step because the frequency spectrum of a 1 point

signal is equal to itself. Although there is no work involved, it is to be

remembered that each of the 1 point signals is now a frequency spectrum, and

not a time domain signal.

3. Synthesize the N frequency spectra into a single frequency spectrum.

Here, the N frequency spectra are combined in the exact reverse order that

the time domain decomposition took place. In other words, this synthesis

must undo the interlaced decomposition done in the time domain.


Gulbarga

15


16/86


17/86


categories or classes. The objects of interest are generically called patterns and

in our case are sequences of acoustic vectors that are extracted from an input

speech using the MFCC technique described in the previous section. The

classes here refer to individual speakers. Since the classification procedure in

our case is applied on extracted features, it can be also referred to as feature

matching.

The state-of-the-art in feature matching techniques used in speaker

recognition includes Dynamic Time Warping (DTW), Hidden Markov Modeling

(HMM), and Vector Quantization (VQ). In this project, the VQ approach will be

used, due to ease of implementation and high accuracy. VQ is a process of

mapping vectors from a large vector space to a finite number of regions in that

space. Each region is called a cluster and can be represented by its center

called a codeword. The collection of all code words is called a codebook.

1.6.3.2 Clusteringthe training vectors:

After the enrollment session, the acoustic vectors extracted from input

speech of a speaker provide a set of training vectors. The next important step is

to build a speaker-specific VQ codebook for this speaker using those training


Gulbarga

17


18/86


vectors. There is a well-known algorithm, namely LBG algorithm [Linde, Buzo

and Gray, 1980], for clustering a set of L training vectors into a set of M

codebook vectors. I have used a simple version of the LBG algorithm, which is

given below:

1. Determine the number of codewords, N or the size of the codebook.

2. Select N codewords at random and let that be the initial codebook. The initial

codewords can be randomly chosen from the set of input vectors.

3. Using the Euclidean distance measure clusterize the vectors around each

codeword.

Do this by taking each input vector and finding the Euclidean distance

between it and each codeword. The input vector belongs to the cluster of the

codeword that yields the minimum distance.

4. Compute the new set of codewords.

Obtaining the average of each cluster does this. Add the component of each

vector and divide by the number of vectors in the cluster as given below:

where i is the component of each vector and m is the number of vectors in the

cluster.

5. Repeat steps 3 and 4 until either the codewords dont change or the change

in the codewords are small.


Gulbarga

18

yi= 1/m x

ij

j=1

(1.6)


19/86


1.6.3.3 Comparison of codebooks:

In the enrollment phase, a speaker-specific VQ codebook is generated for

the speaker by clustering his/her training acoustic vectors. The distance from a

vector to the closest codeword of a codebook is called a VQ-distortion. In the

verification phase, an input utterance of the speaker is vector-quantized using

his trained codebook and the total VQ distortion is computed. If this total VQ

distortion is below the threshold value, then the speaker is authenticated else the

speaker is not authenticated.

1.7 Limitations of the study


Gulbarga

19


20/86


The current Voice Biometric Security for Industrial control is though very

efficient as per the demands and requirements of the present trends, but still it

has got some following limitations:

Its hard for the system to distinguish background noise from the usersvoice. This might affect the systems accuracy.

A persons physical conditions such as sickness or sore throat willaffect the performance. This could lead to the problem of false

rejection in which the right person is rejected as wrong

person/imposter.

A small number of false acceptances can be noticed when the systemaccepts an imposter as a right person to be authenticated.

Our system performs poorly if the voice phrase is very small or if thereis a considerable difference in the enrollment stage and verification

stage. The factors responsible for this difference may be variance in

loudness, environmental conditions or the distance between the

speaker and microphone.

CHAPTER - 2


Gulbarga

20


21/86


SYSTEM ANALYSIS

This chapter gives an analysis of Voice Biometric Security for Industrial

control. This includes project overview, functional requirements, system

requirements, technical specifications, developers responsibility overview and

Use-Case Driven analysis of the system using use-case diagrams.

2.1 Documentation Overview

The document is subdivided into following topics:

Project overview

Functional requirements

System requirements

Technical specifications

Developers responsibility overview

2.1.1 Project overview

The current work is to automate the process of voice authentication. Voice

authentication can be divided into user registration and user verification. The

user registration module is to register the users voice characteristics in the

database file. The user verification module finds the correlation between the input

voice template and the stored voice template in the database.


Gulbarga

21


22/86


2.1.2 Functional requirements User asks to register the voice and the username.

Application receives the request.

The system checks for the existence of the corresponding username in thestored database file. If it is not present, that will be registered in the database file.

User asks for the verification of the voice.

System receives the request and processes the request.

System returns the credibility of the user.

These requirements will be explained by the Use-case diagrams of the UML

standard in the last section of this chapter.

2.1.3 System requirements

A Noise free environment is required for the successful operation of the

proposed system.

2.1.3.1 Hardware Requirements

Processor : Pentium or above

RAM : 32 MB or above

Hard disk : min. of 1 GB ( with at least 30 MB of free space)

Peripheral Devices : SVGA Color Monitor, Mouse , Keyboard


Gulbarga

22


23/86


High Quality Microphone, Speakers.

2.1.3.2 Software Requirements

Operating System : Windows 95/98/Me/Xp/2000 , NT

Software : Visual C++, Sound Card Interface, MSDN Online Help.

2.1.4 Technical specifications

The specifications were drawn up using Object-Oriented Analysis, and

design was constructed using Unified Modeling Language (UML). The whole idea

was implemented in Visual C++.

2.1.5 Developers responsibility overview

The overall objective is to deliver a fault free product on time.

Assumptions:

The voice is taken in a noise free environment through a high qualitymicrophone and is stored in a file. And the file is read for further processing.

The Architecture must be open so that any modules can be added.

The product must be easy to operate and should be user friendly.

The product must conform to the clients hardware.


Gulbarga

23


24/86


And moreover, the client is assumed to be inexperienced with computers.

Special attention was therefore paid to the specification phase and

communication with the client. The product has to be made as user friendly as

possible. There is always the possibility of a major design fault; so extensive

testing must be done during the design phase. The product must meet the

specified storage requirements and response times.


Gulbarga

24


25/86


2.2 Use-Case Driven Analysis

The first step in the analysis is to define the use-cases, which describe what

the system provides in terms of functionality- the functional requirement of the

system. A use-case analysis of UML standard involves reading and analyzing the

specifications as well as discussing the system with potential users of the system.

The next chapter introduces to UML modeling and explains how my system is

modeled using UML.

The use-case diagrams describe the functionality of the system and users

of the system. A use case is a description of a systems behavior from users

standpoint. The main elements of this type of diagrams are: 1.Actorsrepresents

the users of a system including human beings and other systems.

2. Use Cases represent the functionality or services provided by the system to

users.

Now let us see use-case diagrams which depict the project at differentlevels:

(1) The following use-case diagram Fig.2.2 (a) depicts the project at first level,

where the actor is shown as a stick figure with the name of the actor below the

figure and the use case is shown as an ellipse containing the name of the use

case.


Gulbarga

25


26/86


First Level Use-Case diagram:

Description of the diagram:

1. Use Case : User Registration

Actor : User

Desired Outcome : The User name is registered.

Entered When : The user submits his name, clicks on User

Registration and records his voice (some kind

of pass phrase).

Finished When : The users name is registered.

Description : The use-case is to register the user name

when it is not present in the database file.

2. Use Case : User Verification


Gulbarga

26

User Registration

User Verification

User Management

User

Fig.2.2 (a): Use-Case diagram for the proposed system


27/86


Actor : User

Desired Outcome : Is whether the user is authorized or not.


Verification and submits his voice.

Finished When : The result is given.

Description : This use-case is for checking the credibility of

the user by comparing the submitted

(recorded) pass phrase (after converting into

template) with the already existing template in

the database.

3. Use Case : User Management

Actor : User

Desired Outcome : User will be deleted or his / her pass phrase

will be changed according to the requirements.


Management and if required submits his voice.

Finished When : The user Request is processed.

Description : This is an option to the administrator to

Manage the users.


Gulbarga

27


28/86


(2) The following use-case diagrams Fig.2.2 (b), 2.2(c) and 2.2(d) depict the

project at second level, where the actor is shown as a class rectangle with the

label and the use case is shown as an ellipse containing the name of

the use case.

Second Level Use-Case diagrams:

a) Use-Case diagram for User Registration

Fig.2.2 (b): Use-case Diagram for User Registration


Use Case : Enroll

Actor : Registration module

Desired Outcome : Write the template to the database file.

Entered when : The user selects the registration and supplies name, pass

phrase as voice and voice signal is sampled.

Finished When : The characteristics of the voice (template) and

username are written to the database file.


Gulbarga

28

Registration

module

Enroll

Readtemplate file

Generatetemplate

Sample voicesignal

uses

uses

uses


29/86


Description : This use case writes the template to the database file.

b) Use-Case diagram for User Verification

Fig.2.2(c): Use-case Diagram for User Verification


Use Case : Verify

Actor : Verification module

Desired Outcome : Verifies whether valid user or not.

Entered when : The user selects the verification and supplies

name, pass phrase as voice and voice signal is sampled.

Finished When : The final result-accepted or rejected is displayed.

Description : This use case verifies the identity of a user.


Gulbarga

29

Verification

module

Verify

Comparetemplates

Generatetemplate

Sample voicesignal

uses

uses

uses


30/86


c) Use-Case diagram for User Management

Fig.2.2 (d): Use-case diagram for User Management


1. Use Case : Delete User

Actor : Management module

Desired Outcome : The user template is deleted from the file.

Entered when : The user selects the Management module and

submits his name.

Finished when : The user template is deleted from the database

Description : This use case deletes the user template from

the database. It in turn uses the use case Write to file.


Gulbarga

30

Management

module

Change Passphrase

Delete User

Write to file

uses

uses


31/86


2. Use Case : Change Pass phrase

Actor : Management module

Desired Outcome : The modified user template.

Entered when : The user selects the Management module and

submits his name to which the pass phrase has to

be modified.

Finished when : The user template is modified.

Description : This use case modifies the user pass phrase

and stores it back into the database. It inturn

uses the use case Write to file.


Gulbarga

31


32/86


CHAPTER - 3

SYSTEM DESIGN

3.1 Introduction

The Design is the first step into the solution domain. The Voice Biometric

Security for Industrial control is designed into 5 main class modules:

CVoiceBiometricDlg, CSound, CParam CMath, and CVectorQuantization. The

following diagram Fig.3.1 shows these classes used in the system:

Fig.3.1: Classes used in the system


Gulbarga

32

CParam CVectorQuantization

CVoiceBiometricDlg : Enroll CVoiceBiometricDlg : Verify

CVoiceBiometricApp

CSound

CMath


33/86


Description :

The Voice Biometric Security for Industrial control first starts by creating

an instance of the CVoiceBiometricApp object. Then its InitInstance( ) function

creates a dialog window of the type CVoiceBiometricDlg. The

CVoiceBiometricDlg class mainly presents the user interface features and is a

coordination point for calling various functions. When its OnEnroll( ) function is

called, a CSound object is created and pass phrase is recorded in a temp.wav

file. This wave file thus generated by the CSound class is used by the CParam

class for generating MFCCs in a temp.mfc file. Then it is used by

CVectorQuantization class to produce a codebook file. In enrollment phase this

codebook file is saved as username.cb (for example, if the username is

Kanchan, then the codebook file is saved as kanchan.cb). In Verification phase

same steps are repeated and the codebook is saved as temp.cb which is used

by compcbmain() to compare the claimed user codebook with the current

codebook. If the matching score is appropriate user is authenticated else user is

not authenticated.


Gulbarga

33


34/86


3.2 UML Representation Of The System Design

3.2.1. Introduction to UML

The Voice Biometric Security for Industrial control is designed using

Object oriented design methodology (Unified Modeling Language).

Over the past decade, Grady booch, James Rumbaugh and Jacobson

have collaborated to combine the features of their individual object oriented

analysis and design method into a Unified method, the result called the Unified

Modeling Language (UML), has become widely used throughout the industry.

The Unified Modeling Language (UML) is a modeling language for

specifying, visualizing, constructing, and documenting the artifacts of system-

intensive process. UML allows us to express an analysis model using a modeling

notation that is governed by a set of syntactic, semantic and pragmatic rules.

This language unifies the industrys best engineering practices for modeling the

systems. Some worth referring points about UML are:

The UML is not simply a notation for drawing diagrams, but a completelanguage for capturing knowledge about a subject and expressing knowledge

regarding the subject for the purpose of communication.

UML applies to modeling and systems. Modeling involves a focus on understandinga subject and capturing and being able to communicate this knowledge.


Gulbarga

34


35/86


UML is used for specifying, visualizing, constructing, and documentingsystems.

It is based on object-oriented paradigm.

3.2.2 UML Diagrams

The UML metamodel elements are organized into diagrams. Different

diagrams are used for different purposes depending on the angle from which you

are viewing the system. The different views are called architectural views.

Architectural views facilitate the organization of knowledge, and diagrams enable

the communication of knowledge.

The UML defines nine types of diagrams: class, object, use case,

sequence, collaboration, statechart, activity, component, and deployment

diagram. All of these diagrams are based on the principle that concepts are

depicted as symbols and relationships among concepts are depicted as paths

connecting symbols, where both of these types of elements may be named.

Now let us see an architectural view called Behavioral Model View, which

encompasses the dynamic, or behavioral, aspects of a problem and solution. Our

system design can be best explained using the following diagrams of this view:

(1) Sequence Diagrams:


Gulbarga

35


36/86


These diagrams describe the behavior provided by a system to inter-

actors. These diagrams contain classes that exchange messages within an

interaction arranged in time sequence. In generic form, these diagrams describe

a set of message exchange sequences among a set of classes. In instance form,

these diagrams describe how objects of those classes interact with each other

and how messages are sent and receiveThe following Sequence Diagrams

Fig.3.2.2(a), 3.2.2(b) and 3.2.2(c) show the interactions between the objects of

the classes used in the system during different phases. These objects are

represented in usual way as named rectangles and appear on the horizontal axis

of the diagram from left to right. Messages are represented as solid line arrows,

which give the interaction taken place among the objects in a specified

sequence. The time is represented by vertical line, which is a line extending from

the object and is called as the objects lifeline. This lifeline represents the objects

existence during the interaction(a) The following sequence diagram Fig.3.2.2(a)

shows the interactions between the objects of the classes used in the system

during enrollment phase:


Gulbarga

36


37/86


temp.cb

Fig.3.2.2(a): Sequence diagram for User Enrollment


Gulbarga

37

CVoiceBiometricDlg Enrollment CSound CParam CVectorQuantization

username.cb

temp.wav

temp.wav

temp.mfc

temp.mfc

RecordStart()OnEnroll()


38/86


Description :

1. When the user tries to register his voice in the database, an OnEnroll( )

function of the CVoiceBiometricDlg object is called which inturn calls the

RecordStart( ) function of the object CSound, to record the user voice.

2. After recording, the voice is stored into the temp.wav file, which is then

used by CParam object to extract the voice features from it.

3. After extraction, the extracted features are stored into the temp.mfc file,

which is then used by CVectorQuantization object, to generate the

temp.cb codebook file.

4. The generated codebook file is then saved as username.cb into the

database by the function saveusercb( ) of CVoiceBiometricDlg object.


Gulbarga

38


39/86


(b) The following sequence diagram Fig.3.2.2(b) shows the interactions between

the objects of the classes used in the system during verificationphase:

OnVerify( )

compcbmain( )

Fig.3.2.2(b): Sequence diagram for User Verification


Gulbarga

39

CVoiceBiometricDlg Verification CSound CParam CVectorQuantization

Result

temp.wav

temp.wav

temp.mfc

temp.mfc

RecordStart( )

temp.cb


40/86


Description :

1. When the user tries to verify his registered voice, an OnVerify( ) function

of the CVoiceBiometricDlg object is called which inturn calls the


2. The steps 2 and 3 of Fig.3.2.2(a) are followed similarly.

3. The generated temp.cb codebook is then compared with the claimed

username.cb codebook by the function compcbmain( ) of

CVectorQuantization object.

4. Depending on the matching score, the CVoiceBiometricDlg object

declares the Result to the user as he is authenticated or he is not

authenticated.


Gulbarga

40


41/86


(c) The following sequence diagram Fig.3.2.2(c) shows the interactions between

the objects of the classes used in the system when the user modify his

voice:

OnModify( )

temp.cb

Fig.3.2.2(c): Sequence diagram for User Modification


Gulbarga

41

CVoiceBiometricDlg Modification CSound CParam CVectorQuantization

temp.wav

temp.wav

temp.mfc

temp.mfc

Replace oldusercb withnew usercb

RecordStart( )


42/86


Description :

1. When the user tries to modify his registered voice, an OnModify( ) function

of the CVoiceBiometricDlg object is called which inturn calls the


2. The steps 2 and 3 of Fig. 3.2.2(a) are followed similarly.

3. The old username.cb is then replaced with the new username.cb

codebook.

(2) Statechart Diagrams:

These diagrams render the states and responses of a class participating

in behavior, and the life cycle of an object. These diagrams describe the behavior

of a class in response to external stimuli.

In the following statechart diagrams Fig.3.2.2(d), 3.2.2(e) and 3.2.2(f) for

the different classes of the system, the state is represented as a rounded box

that describes the class at a specific point in time. The states shown in all these 3

diagrams are self-explanatory. The initial state is shown as a solid circle, and the

final state is shown as a circle surrounding a small solid circle, a bulls-eye. The

transition is represented as a solid line arrow, which shows the relationship


Gulbarga

42


43/86


between the two states indicating that the class in the first state will enter the

second state and perform certain actions when a specific event occurs.

(a)The following diagram Fig.3.2.2(d) shows the Statechart diagram for CSound

class in the system:


Gulbarga

43

Loads the defaultrecording format

Begin recordingvoice

Gets the givenrecording format

Stop recording voice

Start

Stop

Fig. 3.2.2(d): Statechart diagram for CSound class


44/86


(b) The following diagram Fig.3.2.2 (e) shows the Statechart diagram for CParam

class in the system:


Gulbarga

44

Stop

Continuous speech signalis divided into frames Each frame is

windowed

FFT is applied to eachwindowed frame

Spectrum coeff. areconverted into melfrequency spectrum coeff.

Cepstrum is applied toeach mel spectrum, we getMFCC coeff.

Start


45/86


Fig.3.2.2 (e): Statechart diagram for CParam class


Gulbarga

45

Stop


46/86


(c) The following diagram Fig.3.2.2(f) shows the Statechart diagram for

CVectorQuantization class in the system:

Fig.3.2.2 (f): Statechart diagram for CVectorQuantization class


Gulbarga

46

Initialize the size ofthe codebook

By applying VQ algm.,clusterize the vectorsto prepare codebook

Save the codebookin a file

Starttt

Stop

Compare this codebook with the trainedcodebook to calculate totalVQ distortion

for Verification phase


47/86


(3) Activity Diagram:

Activity diagrams render the activities or actions of a class participating in

behavior. These diagrams describe the behavior of a class in response to

internal processing rather than external events. Activity diagrams describe the

processing activities within a class.

An activity diagram is similar to a statechart diagram, where states are

activities (rounded boxes) representing the performance of operations and the

transitions (solid line arrows) are triggered by the completion of the operations.

The following activity diagram Fig.3.2.2 (g) shows the activities of 4 main

phases of the system: Enroll, Modify, Verify & Delete (shown from left to right).

All the activities shown in this diagram are self-explanatory.


Gulbarga

47

stop


48/86


Fig.3.2.2 (g): Activity diagram of the system


Gulbarga

48

User has to enter

his name

User is asked for

new pass phrase

User is asked for

pass phrase

User is deleted

from the database

User is asked for

pass phrase

MFCC coeff. are

extracted from the

recorded pass phrase

By applying vector

quantization algm.,

clusterize the

vectors

Compare these features

with already stored

features in the file

Voice featuresare stored in a file

MFCC coeff. are

extracted from the


By applying vector

quantization algm.,

clusterize the vectors

Voice features arestored in a file which

is

then replaced with the

old file

MFCC coeff. are

extracted from the


By applying vector

quantization algm.,

clusterize the vectors

Voice featuresare stored in a file

Message will be

displayed for

real user/imposter

Start

Stop


49/86


CHAPTER - 4

IMPLEMENTATION

4.1 Coding Details

During the course of my involvement with speaker verification, I have

come across several papers with information about speaker verification

technologies but unfortunately most of them were not concerned with the topic.

Many of them represent the theoretical aspects but very few present the

implementation methodologies.

I have implemented a Voice Biometric Security for Industrial control that

uses Mel Frequency Cepstrum Coefficients ( MFCC ) which are used for feature

extraction and Vector Quantization ( VQ ) for feature matching.

For sound recording, the duration of recording has been set to three

seconds. The exact timing of an utterance will generally not be the same as that

of the template. If a person speaks just a small word (e.g. Hello, or any small

word), he may not be authenticated although he is a genuine user and other may

easily login to his account. For this purpose if the utterance length is less than

one second the system will ask to speak a longer phrase.


Gulbarga

49


50/86


The genuine users can be correctly authenticated, provided that they

speak in the same way (including the emotions, loudness, and bearing in mind

that this is a word-dependent and speaker dependent system) as they did at the

time of the enrollment. The authentication becomes easier if the recording time

for the speech is at the maximum of two seconds.

The system is implemented in object-oriented paradigm. The system has

been divided into 5 main classes consisting of CSound for recording wave files,

CParam for extracting the MFCCs from the wave file, CVectorQuantization for

generating the code book from the MFCCs. CMath is a class which provides an

easy way for calculating Fast Fourier Transform etc. CVoiceBiometricDlg was the

main dialog where the user interface is implemented. The other classes in the

system were of various controls and of some ActiveX controls. The next section

explains the various functions used by these classes and how the development

of the code is being proceeded.


Gulbarga

50


51/86


4.2 Detail Implementation

The following Table 4.2 lists the main functions used by the classes in the

system:

CLASS FUNCTION

CvoiceBiometricDlg OnEnroll( ), OnVerify( ) , OnModify( ),

OnDelete( ), registeruser( ), SetTimer( ),

OnTimer( ), SaveFile( ), RemoveSilence( )

rkmain( ), WaveLoad( ), vq( ),

saveusercb( )

Csound RecordStart( ), RecordStop( )

Cparam InitFBank( ), Mel( ), InitMFCC( ),

wave2MFCC( ), DoHamming( )

UnInitMFCC( ), UnInitFBank( )

Cmath fft( ), fftsort( ), dct( )

CvectorQuantization initcb( ), clusterize( ), eucldist( ),

writecb( ), compcbmain( )

Table 4.2: Classes and their functions used in the system


Gulbarga

51


52/86


CVoiceBiometricDlg class provides four important functions to manipulate

the users voice signal. These are:

1. OnEnroll( ) : It registers the users voice features in the database.

2. OnVerify( ) : It verifies the current users voice features with the claimed

users voice features.

3. OnModify( ) : It replaces the old users voice features with the new users

voice features.

4. OnDelete( ) : It just deletes the users voice features from the database.

The above first three functions inturn use various functions of the other

four classes- CSound, CParam, CMath, and CVectorQuantization, to accomplish

their tasks. Let us see how these functions proceed during the development of

code:

(1) OnEnroll( ) function:

The steps followed by this function are as follows:

1. The users voice is recorded with the help of registeruser( ) function which

inturn takes the help of following functions:

a) RecordStart( ) : To begin the recording of users voice.


Gulbarga

52


53/86


b) SetTimer( ) : Its a function of a Microsoft Foundation Class (MFC)

called CWnd. It is used to set the timer for 3 seconds

to record the users voice.

2. When the timer elapses, the function OnTimer( ) is called in which the

timer code is written. In this code, all of the following steps are carried out.

3. The RecordStop( ) function is called to stop the recording.

4. Recorded voice is stored into a file called temp.wav by the function

SaveFile( ).

5. The RemoveSilence( ) function is called to remove the silence present in

temp.wav file.

6. The rkmain( ) function is called to extract the MFCCs from the temp.wav

file. For this it uses the following steps:

a) First, temp.wav file is loaded into the memory by the function

WaveLoad( ).

b) Then all the functions of CParam class are called in the following

order:

i) InitFBank( ) : To initialize the FBankInfo structure which gives

the information needed about the filter banks to be created. It

inturn uses Mel( ) to apply mel-frequency scale on each frame.

ii) InitMFCC( ) : To initialize the MFCCInfo structure which gives

the information needed about the MFCCs to be created.


Gulbarga

53


54/86


iii) wave2MFCC( ) : To extract the MFCCs from temp.wav file. The

MFCCs are stored into a file called temp.mfc.This function inturn

calls other functions in the following order:

DoHamming( ) to apply windowing on each frame.

fft( ) to apply Fast Fourier Transform algorithm.It inturncalls fftsort( ) to implement Interlaced decomposition by

using Bit reversal sorting algorithm which involves

rearranging the order of the N time domain samples by

counting in binary with the bits flipped left-for-right.

dct( ) to apply Discrete Cosine Transform on each frame.iv) UnInitMFCC( ) : To uninitialize the MFCCInfo when the work

is done.

v) UnInitFBank( ) : To uninitialize the FBankInfo when the work

is done.

7. The vq( ) function is called to clusterize the acoustic vectors of temp.mfc into

a codebook called temp.cb. It inturn calls the other functions of class

CVectorQuantization in the following order:

a) initcb( ) : To initialize the size of the codebook.

b) clusterize( ) : To clusterize the acoustic vectors using LBG algorithm. It

inturn uses the function eucldist( ) to calculate the Euclidean distance

between the vectors.

c) writecb( ) : To store the codebook in a file called temp.cb.


Gulbarga

54


55/86


8. The file temp.cb is stored as username.cb into the database by the function

saveusercb( ).

(2) OnVerify( ) function:

The steps followed by this function are as follows:

1. The steps 1 to 7 used by the OnEnroll( ) function are followed similarly,to

generate the codebook temp.cb for the current user.

2. To verify the current users codebook with his trained codebook, the

compcbmain( ) function is then called, where the total VQ distortion is

calculated by comparing current temp.cb with claimed username.cb of the

user.

(3) OnModify( ) function:

The steps 1 to 8 used by the OnEnroll( ) function are followed similarly,but in

the step 8, the old username.cb is replaced with the new temp.cb of the user.


Gulbarga

55


56/86


CHAPTER - 5

Output


Gulbarga

56


57/86


CHAPTER - 6

Hardware Specifications

Parallel port modes

The IEEE 1284 Standard which has been published in 1994 defines five

modes of data transfer for parallel port. They are:

1. Compatibility Mode

2. Nibble Mode

3. Byte Mode

4. EPP

5. ECP


Gulbarga

57


58/86


As the name refers, data is transferred over data lines. Control lines are

used to control the peripheral, and of course, the peripheral returns status signals

back to the computer through Status lines. These lines are connected to Data,

Control And Status registers internally. The details of parallel port signal lines are

given below:

Pin No(DB25)

Signal name Direction Register bit Inverted

1 nStrobe Out Control-0 Yes

2 Data0 In/Out Data-0 No








10 nAck In Status-6 No

11 Busy In Status-7 Yes

12 Paper-Out In Status-5 No

13 Select In Status-4 No

14 Linefeed Out Control-1 Yes

15 nError In Status-3 No

16 nInitialize Out Control-2 No

17 nSelect-Printer Out Control-3 Yes

18-25 Ground - - -

6.1 Parallel port registers

As you know, the Data, Control and Status lines are connected to there

corresponding registers inside the computer. So, by manipulating these registers


Gulbarga

58


59/86


in program, one can easily read or write to parallel port with programming

languages like 'C' and BASIC.

The registers found in a standard parallel port are:

1. Data register

2. Status register

3. Control register

As their names specify, Data register is connected to Data lines, Control

register is connected to Control lines and Status register is connected to Status

lines. (Here the word connection does not mean that there is some physical

connection between data/control/status lines. The registers are virtually

connected to the corresponding lines.) So, whatever you write to these registers

will appear in the corresponding lines as voltages. Of course, you can measure it

with a multimeter. And whatever you give to Parallel port as voltages can be read

from these registers (with some restrictions). For example, if we write '1' to Data

register, the line Data0 will be driven to +5v. Just like this, we can

programmatically turn on and off any of the Data lines and Control lines.

6.2 Where these registers are?

In an IBM PC, these registers are IO mapped and will have an unique

address. We have to find these addresses to work with the parallel port. For a

typical PC, the base address of LPT1 is 0x378 and of LPT2 is 0x278. The Data

register resides at this base address, Status register at base address + 1 and the


Gulbarga

59


60/86


control register is at base address + 2. So, once we have the base address, we

can calculate the address of each register in this manner. The table below shows

the register addresses of LPT1 and LPT2.

Register LPT1 LPT2

Data register (baseaddress + 0) 0x378 0x278

Status register (baseaddress + 1) 0x379 0x279

Control register (baseaddress + 2) 0x37a 0x27a

Standard LCD Pin Matches (Character number


61/86


4 RS 0/10 = Instructioninput / 1 = Datainput

5 R/W 0/10 = Write toLCD module /1 = Read fromLCD module

6 E 1, 1-->0 Enable signal

7 DB0 0/1 Data pin 0



10 DB3 0/1 Data pin 311 DB4 0/1 Data pin 4




If your LCD has more than 80 characters (like 4x40)

15 E2 1, 1->0 Enable signal row 2 & 3

16 Not used mostly

HY1602B (Hyper 1602B) with KS0065 controller (compatible with HD44780) and

backlight. Click for a bigger picture


Gulbarga

61


62/86


1602-04 with KS0066U controller (compatible with HD44780) and backlight Click for

a bigger picture

These are 2 different 2x16 LCDs as in the pictures the 15th and 16th pins are for

backlight as I mentioned above.

Pinout Descriptions

Pin 1, 2, 3 :According to the table, I call Pin 1(Vss),2(Vdd/Vcc),3(Vee/Vo) power

pins because they are the gates to power. Pin 1 is for ground so you have to

connect to ground/earth and Pin 2 is for the +5V power supply. 6V or 4,5V is

mostly acceptable in few amperes and also 3V is acceptable on some of the LCD

modules (You can also power these modules with a battery in a very economical

way). In my application I get the voltage from the molex cable of the pc which is


Gulbarga

62
http://www.izdir.com/electronic_work/article2/113_1304.JPGhttp://www.izdir.com/electronic_work/article2/112_1214.JPGhttp://www.izdir.com/electronic_work/article2/111_1133.JPGhttp://www.izdir.com/electronic_work/article2/113_1304.JPG


63/86


inside the case. And pin 3 is for the LCD's contrast adjustment. I did not but you

could use a potentiometer (10K pot will be ok) for changing the contrast of your

LCD. See the schematics below

Pin 4, 5, 6: I call Pin4 (RS), 5(R/W),6(E) the control buddies because these pins

are the arms of your controller inside your LCD module. Pin 4(RS) is registration

select if this pin is low the data perceived by the data pins taken as commands

by the LCD and if this pin is high the LCD can receive/send 8 or 4 bit character

data. I call Pin 5(R/W) clerk because when this pin is low you can write character

to the LCD, if the pin is high you can read character data or the status

information from the LCD. I didn't make any read operations in my app so I solder

this pin to the ground (with soldering this to the ground I made this pin low - "0"see the below circuits). Pin 6(E) which I call the guardian, is used to initiate the

actual transfer of commands or character data between the LCD module and the

data pins.

Pin 7,8,9,10,11,12,13,14: The eight pins which are DB0-DB7 are the data pins

which I call them the workers. The data can be transferred or fetched from the

LCD by 8 or 4 bits. Which one is better? This is up to you, by the way if you are

using a microcontroller and you have few pins you can use your module in 4 bit

mode (by using DB4-DB7). I used 8 bit mode in my LCD because I used the

parallel port which already have 8 bit data lines (remembermy first article, part 1

D0-D7)

Pin 15,16 :These two pins are for the backlight of the LCD module. 15th pin goes

to the power supply(VB+) and 16th pin goes to the ground(VB-). Backlight is very

useful in dim environments but some LCD modules don't have backlights. There

are multicolored LCDs around as well.

Circuit


Gulbarga

63
http://www.codeproject.com/csharp/csppleds.asphttp://www.codeproject.com/csharp/csppleds.asp


64/86


What we need to supply for our circuit? Below is a list of that:

No: Description

1. 2x16 Paralell LCD must be HD44780 compatiable

2. Normal parallel printer cable3. Normal twin power cable

4. 10 way housing and the PCB header

5. 6 way housing and the PCB header

6. Hard drive type power connector (Molex)

7. 16 PCB terminals

8.A digital multimeter(must measure few ampers!), a solder penwith some soldering iron

9. Some patience, and my program :)

*10 K potentiometer (Not required - needed when you want toadjust the contrast of your LCD - see circuit withpotentiometer)

The circuit with potentiometer


Gulbarga

64
http://www.izdir.com/electronic_work/article2/112_1297.JPGhttp://www.izdir.com/electronic_work/article2/112_1298.JPGhttp://www.izdir.com/electronic_work/article2/112_1300.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPGhttp://www.izdir.com/electronic_work/article2/113_1304.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPGhttp://www.izdir.com/electronic_work/article2/112_1297.JPGhttp://www.izdir.com/electronic_work/article2/112_1298.JPGhttp://www.izdir.com/electronic_work/article2/112_1300.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPGhttp://www.izdir.com/electronic_work/article2/113_1304.JPGhttp://www.izdir.com/electronic_work/article2/112_1299.JPG


65/86


If your soldering goes well you get a typical test screen of a character based LCD

as its shown below:


Gulbarga

65


66/86


Power Supply

Regulated power supply:


Gulbarga

66


67/86


68/86


LM7805 5 1.5 10 3 2

LM7806 6 1.5 12 5 2

Hardware:

Everybody knows what is parallel port, where it can be found,

and for what it is being used. The primary use of parallel port is to

connect printers to the computer and is specifically designed for this

purpose. Thus it is often called as printer Port or Centronics port (this

name came from a popular printer manufacturing company

'Centronics' which devised some standards for parallel port). You can

see the parallel port connector in the rear panel of your PC. It is a 25

pin female (DB25) connector (to which printer is connected). On almost

all the PCs only one parallel port is present, but you can add more by

buying and inserting ISA/PCI parallel port cards.

The power supply is 9 volt regulated power supply. We have used a step down

transformer of 9 volt. The result is rectified using a bridge rectifier circuit. The

result is passed through regulator IC and the Out put is coupled through a 1

microfarad capacitor for better impedence matching.

STEPPING MOTORS:


Gulbarga

68


69/86


Motion Control, in electronic terms, means to accurately control

the movement of an object based on speed, distance, load, inertia or a

combination of all these factors. There are numerous types of motion

control systems, including; Stepper Motor, Linear Step Motor, DC

Brush, Brushless, Servo, Brushless Servo and more. This document will

concentrate on Step Motor technology.

In Theory, a Stepper motor is a marvel in simplicity. It has no

brushes, or contacts. Basically it's a synchronous motor with the

magnetic field electronically switched to rotate the armature magnet

around.

A Stepping Motor System consists of three basic elements, often

combined with some type of user interface (Host Computer, PLC or

Dumb Terminal):

The Indexer (or Controller) is a microprocessor capable of generating step pulses

and direction signals for the driver. In addition, the indexer is typically required to

perform many other sophisticated command functions.

The Driver (or Amplifier) converts the indexer command signals into the powernecessary to energize the motor windings. There are numerous types of drivers, with

different current/amperage ratings and construction technology. Not all drivers are

suitable to run all motors, so when designing a Motion Control System the driver

selection process is critical. The Step Motor is an electromagnetic device that convertsdigital pulses into mechanical shaft rotation.


Gulbarga

69


70/86


Advantages of step motors are low cost, high reliability, high

torque at low speeds and a simple, rugged construction that operates

in almost any environment. The main disadvantages in using a step

motor is the resonance effect often exhibited at low speeds and

decreasing torque with increasing speed.

TYPES OF STEPPER MOTORS:

There are basically three types of stepping motors; variable

reluctance, permanent magnet and hybrid.

They differ in terms of construction based on the use of permanent magnets

and/or iron rotors with laminated steel stators.

VARIABLE RELUCTANCE:

The variable reluctance motor does not use a permanent

magnet. As a result, the motor rotor can move without constraint or

"detent" torque. This type of construction is good in non industrial

applications that do not require a high degree of motor torque, such as

the positioning of a micro slide.

The variable reluctance motor in the below illustration has four "stator pole sets"

(A, B, C,), set 15 degrees apart. Current applied to pole A through the motor winding


Gulbarga

70


71/86


causes a magnetic attraction that aligns the rotor (tooth) to pole A. Energizing stator poleB causes the rotor to rotate 15 degrees in alignment with pole B. This process will

continue with pole C and back to A in a clockwise direction. Reversing the procedure (C

to A) would result in a counterclockwise rotation.

FIG: VARIABLE RELUCTANCE

MOTOR

PERMANENT MAGNET:

The permanent magnet motor, also referred to as a "canstack"

motor, has, as the name implies, a permanent magnet rotor. It is a

relatively low speed, low torque device with large step angles of either

45 or 90 degrees. It's simple construction and low cost make it an ideal

choice for non industrial applications, such as a line printer print wheel

positioner.


Gulbarga

71


72/86


Fig: Permanent Magnet Motor

Unlike the other stepping motors, the PM motor rotor has no

teeth and is designed to be magnetized at a right angle to it's axis. The

above illustration shows a simple, 90 degree PM motor with four

phases (A-D). Applying current to each phase in sequence will cause

the rotor to rotate by adjusting to the changing magnetic fields.

Although it operates at fairly low speed the PM motor has a relatively

high torque characteristic.

HYBRID:

Hybrid motors combine the best characteristics of the variable

reluctance and permanent magnet motors. They are constructed with

multi-toothed stator poles and a permanent magnet rotor. Standard

hybrid motors have 200 rotor teeth and rotate at 1.80 step angles.

Other hybrid motors are available in 0.9and 3.6 step angle

configurations. Because they exhibit high static and dynamic torque


Gulbarga

72


73/86


74/86


Fig: Lead Biflar MotorThe most common wiring configuration for bifilar wound stepping

motors is 8 leads because they offer the flexibility of either a Series or

parallel connection. There are however, many 6 lead stepping motors

available for Series connection applications.

STEP MODES:

Stepper motor "step modes" include Full, Half and Microstep. The

type of step mode output of any motor is dependent on the design of

the driver.

FULL STEP:

Standard (hybrid) stepping motors have 200 rotor teeth, or 200

full steps per revolution of the motor shaft. Dividing the 200 steps into

the 360's rotation equals a 1.8 full step angle. Normally, full step

mode is achieved by energizing both windings while reversing the


Gulbarga

74


75/86


current alternately. Essentially one digital input from the driver is

equivalent to one step.

HALF STEP:

Half step simply means that the motor is rotating at 400 steps

per revolution. In this mode, one winding is energized and then two

windings are energized alternately, causing the rotor to rotate at half

the distance, or 0.9's. (The same effect can be achieved by operating

in full step mode with a 400 step per revolution motor). Half stepping is

a more practical solution however, in industrial applications. Although

it provides slightly less torque, half step mode reduces the amount

"jumpiness" inherent in running in a full step mode.

MICROSTEP:

Microstepping is a relatively new stepper motor technology that

controls the current in the motor winding to a degree that further

subdivides the number of positions between poles. AMS microsteppers

are capable of rotating at 1/256 of a step (per step), or over 50,000

steps per revolution.


Gulbarga

75


76/86


FIG: RESOLUTION VS CPU STEP FREQUENCY

Microstepping is typically used in applications that require

accurate positioning and a fine resolution over a wide range of speeds.

MAX-2000 microsteppers integrate state-of-the-art hardware with

"VRMC" (Variable Resolution Microstep Control) technology developed

by AMS. At slow shaft speeds, VRMCs produces high resolution

microstep positioning for silent, resonance-free operation. As shaft

speed increases, the output step resolution is expanded using "on-

motor-pole" synchronization. At the completion of a coarse index, the

target micro position is trimmed to 1/100 of a (command) step to

achieve and maintain precise positioning.

DESIGN CONSIDERATIONS:


Gulbarga

76


77/86


The electrical compatibility between the motor and the driver are

the most critical factors in a stepper motor system design. Some

general guidelines in the selection of these components are:

INDUCTANCE:

Stepper motors are rated with a varying degree of inductance. A

high inductance motor will provide a greater amount of torque at low

speeds and similarly the reverse is true.

SERIES, PARALLEL CONNECTION:

There are two ways to connect a stepper motor; in series or in

parallel. A series connection provides a high inductance and therefore

greater performance at low speeds. A parallel connection will lower the

inductance but increase the torque at faster speeds. The following is a

typical speed/torque curve for an AMS driver and motor connected in

series and parallel:


Gulbarga

77


78/86


Fig: Torque versus speed for series and parallel connection

DRIVER VOLTAGE:

The higher the output voltage from the driver, the higher the

level of torque vs. speed. Generally, the driver output voltage should

be rated higher than the motor voltage rating.

MOTOR STIFFNESS:

By design, stepping motors tend to run stiff. Reducing the

current flow to the motor by a small percentage will smooth the

rotation. Likewise, increasing the motor current will increase the

stiffness but will also provide more torque. Trade-offs between speed,

torque and resolution are a main consideration in designing a step

motor system.

MOTOR HEAT:

Step motors are designed to run hot (50-90 C). However, too

much current may cause excessive heating and damage to the motor


Gulbarga

78


79/86


insulation and windings. AMS step motor products reduce the risk of

overheating by providing a programmable Run/Hold current feature.

C#.NET:

.NET (dot-net) is the name Microsoft gives to its

general vision of the future of computing, the view

being of a world in which many applications run in a

distributed manner across the Internet. We can

identify a number of different motivations driving

this vision.

Firstly, distributed computing is rather like object oriented

progamming, in that it encourages specialized code to be collected in

one place, rather than copied redundantly in lots of places. There are

thus potential efficiency gains to be made in moving to the distributed

model.

Secondly, by collecting specialized code in one place and

opening up a generally accessible interface to it, different types of

machines (phones, handheld, desktops, etc) can all be supported with

the same code. Hence Microsofts runanywhere aspiration.


Gulbarga

79


80/86


Thirdly, by controlling real-time access to some of the distributed

nodes (especially those concerning authentication), companies like

Microsoft can control more easily the running of its applications. It

moves applications further in the area of services provided rather

than object owned.

Interestingly, in taking on the .NET vision, Microsoft seems to

have given up some of this proprietary tendencies (whereby all the

tehcnology it touched was warped towards its Windows operating

system). Because it sees its future as providing software services in

distributed applications, the .NET framework has been written so that

applications on other platforms will be able to access these services.

At the development end of the .NET vision is the .NET framework.

This contains the common language runtime, the common language

runtime (CLR) manages the execution of code compiled for the .NET

platform. The CLR has tow interesting features. Firstly, its specification

has been opened up so that it can be ported to non-windows platforms.

Secondly, any number of different languages can be used to

manipulate the .NET framework classes, and the CLR will support

them. This has led one commentator to claim that under.NET the

language one uses is a lifestyle choice.


Gulbarga

80


81/86


Not all of the supported languages fit entirely neatly into the

.NET framework, however (in some cases the fit has been somewhat

procrustean). But the one language that is guaranteed to fit in

perfectly is C#. This new language, a successor to C++, has been

released in conjunction with the .NET framework, and is likely to be the

language of choice for many developers working on .NET applications.


Gulbarga

81


82/86


CHAPTER - 7

CONCLUSION

The aim of this research was to develop a Voice Biometric Security for

Industrial control, which can verify a persons claimed identity. Overall, the

project was a success with the basic requirements being satisfied. The finished

product could enroll users, verify their voiceprint, and provided a GUI interface for

users to do so.

The current Voice Biometric Security for Industrial control verifies the

voice templates in a one-to-one manner and offers a reliable and accurate way of

verifying the user voice. Simple algorithms for Mel Frequency Cepstrum

Coefficients and Vector Quantization were used for the development of the

system.

The performance of the system was found to depend on the recording

time and length of the sentence, sampling frequency, surrounding environment,

speakers behavioral conditions and the threshold value. To increase the length

of the time, the recording time has to be increased. So, for a maximum of 2

seconds, a user was able to record only a short phrase like VTU University. It

was initially hard for the system to authenticate the users with this kind of voice

phrases. So tests have been performed on increasing the recording time. As the


Gulbarga

82


83/86


recording time increases (3-5 seconds), the systems verification process can

authenticate more accurately, reducing the number of false rejections. The

sampling frequency can also be changed from 16 KHz to 22 KHz. The best

results can be observed at 22 KHz of sampling frequency.

The performance of the system can be improved in several ways. From

the literature I have gone through, it is specified that normalization techniques

can be used to alleviate the noise effects. In future I am planning to study

normalization process and implement it in my system. On

voice biometric report

Documents