Scribd Logo
Upload

Welcome to Scribd!

  • 36 views

    Speaker Recognition System: by Divya V

    Uploaded by

    profmns
    This document discusses speaker recognition systems which can verify a speaker's identity or identify an unknown speaker from a group of known speakers. It describes applications such as banking by telephone, telephone shopping, and computer access security. The key components of speaker recognition systems are signal processing, feature extraction, comparison of extracted features to reference patterns, and a decision logic module. There are two main subareas: speaker verification involves a single comparison to verify a claimed identity, while speaker identification involves multiple comparisons to determine the unknown speaker's identity from a set of possibilities.

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd

    Speaker Recognition System: by Divya V

    Uploaded by

    profmns
    36 views13 pages
    This document discusses speaker recognition systems which can verify a speaker's identity or identify an unknown speaker from a group of known speakers. It describes applications such as banking by telephone, telephone shopping, and computer access security. The key components of speaker recognition systems are signal processing, feature extraction, comparison of extracted features to reference patterns, and a decision logic module. There are two main subareas: speaker verification involves a single comparison to verify a claimed identity, while speaker identification involves multiple comparisons to determine the unknown speaker's identity from a set of possibilities.

    Original Description:

    e

    Original Title

    9.2

    Copyright

    © © All Rights Reserved

    Available Formats

    PPT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses speaker recognition systems which can verify a speaker's identity or identify an unknown speaker from a group of known speakers. It describes applications such as banking by telephone, telephone shopping, and computer access security. The key components of speaker recognition systems are signal processing, feature extraction, comparison of extracted features to reference patterns, and a decision logic module. There are two main subareas: speaker verification involves a single comparison to verify a claimed identity, while speaker identification involves multiple comparisons to determine the unknown speaker's identity from a set of possibilities.

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    36 views13 pages

    Speaker Recognition System: by Divya V

    Uploaded by

    profmns
    This document discusses speaker recognition systems which can verify a speaker's identity or identify an unknown speaker from a group of known speakers. It describes applications such as banking by telephone, telephone shopping, and computer access security. The key components of speaker recognition systems are signal processing, feature extraction, comparison of extracted features to reference patterns, and a decision logic module. There are two main subareas: speaker verification involves a single comparison to verify a claimed identity, while speaker identification involves multiple comparisons to determine the unknown speaker's identity from a set of possibilities.

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    You are on page 1of 13

    SPEAKER RECOGNITION

    SYSTEM
    By Divya V
    INTRODUCTION
    To verify a speaker identity or identify the speaker from a
    known ensemble

    APPLICATIONS:

    banking by telephone
    telephone shopping
    database access services
    information services
    voice mail
    security control for confidential information areas
    remote access to computers.



    GENERAL REPRESENTION
    REFERENCE
    PATTERNS
    SIGNAL
    PROCESSOR
    COMPARSION
    DECISION
    LOGIC
    X
    PATTERN
    VECTOR
    D
    DISTANCE
    IDENTIFICATION
    S(n)
    SUBAREAS OF SPEECH
    RECOGINITION
    SPEAKER
    VERIFICATION
    Identity is claimed by
    user

    Single comparison
    between a set of
    measurement and
    reference
    SPEAKER
    IDENTIFICATION
    Identification among
    N speaker in user
    population is made
    N comparison
    between a set of
    measurement and
    reference
    SPEAKER VERIFICATION
    STORED
    VOICE
    PATTERNS
    COMPARE
    DECIDE
    ACCEPT
    OR
    REJECT
    CLAIMED
    IDENTITY
    SPOKEN
    PHRASES
    TRANSACTION
    REQUESTED
    COMPUTER
    SPEECH
    ANSWER BACK
    CUSTOMER
    TO BE
    VERIFIED
    SIGNAL PROCESSING ASPECT OF
    SPEAKER VERIFICATION
    LPF
    LPF
    ENDPOINT
    DETECTOR
    PITCH
    DETECTOR
    ENERGY
    MEASUREMENT
    LPC
    ANALYSIS
    FORMANT
    ANALYSIS
    DYNAMIC

    TIME

    WARPING

    TO

    REGISTER

    PATTERNS
    COMPUTE
    DISTANCE
    ACCEPT
    REJECT
    S(N)
    STORED
    REFERENCE
    PATTERN
    TIME WRAPING
    Pitch, intensity and formant variation of
    speaker is not same all the time
    Hence warping of time scale t of
    reference utterance is done.
    =t+q(t)
    WARPING FUNCTION
    For linear warping:

    Let the points in
    measured contour be n=1,2.N
    reference contour be m=1,2.M
    We wish to choose a time warping function
    m=w(n)

    The boundary condition on w(n) are:
    w(1)=1 beginning point
    w(N)=M ending point

    Then time warping function

    w(n)=[(M-1/N-1)(n-1)+1]



    WARPING FUNCTION
    For non linear warping function:
    To limit the degree of nonlinearity, the warping function cannot change by more
    than 2 grids at any index n
    Thus
    w(n+1)-w(n)=0,1,2 if w(n) w(n-1) if at previous grid index the warped
    index changed
    =1,2 if w(n)= w(n-1) if at previous grid index the warped
    index stayed constant
    Similarity measure between reference at grid index n and test at grid index m is
    measured
    Similarity measure used to determine the path of warping function , locally
    minimizes the maximum total distance
    COMPUTING DISTANCE
    To compare the overall distance to an appropriately
    chosen threshold
    The contour distance measure is
    d
    j
    = [(a
    js
    (i) a
    jr
    (i))/ ]
    2


    a
    js
    (i)-value of the jth measurement counter at time i
    a
    jr
    (i)-value of the jth reference counter at time i
    -standard deviation of jth measurement at time I
    The overall distance D is
    D= w
    j
    d
    j

    w
    j
    - jth weight chosen on basis of effectiveness of jth
    measurement in verifying the speaker



    i
    j
    SPEAKER IDENTIFICATION
    Processing similar to speaker verification
    N distance measurements are made
    Final decision for speaker identification is
    to choose the speaker whose reference
    patterns is closest in distance to sample
    pattern

    DISTANCE MEASUREMENT BY ATAL
    Let x be an L dimensional column vector representing the input pattern
    Assumption:
    It is joint probability density function of measurements for i
    th
    speaker is multi
    dimensional gaussian distribution with m
    i
    and covariance w
    i


    g
    i
    (x)=(2)
    -1/2
    w
    i
    -1/2
    exp[-1/2(x- m
    i
    )
    t
    w
    i
    -1
    (x- m
    i
    )]
    Decision rule for speaker identification to minimize the probability of error
    p
    i
    g
    i
    p
    j
    g
    j
    for all i j
    Decision class i if
    d
    i
    (x)= 1/2(x- m
    i
    )
    t
    w
    i
    -1
    (x- m
    i
    )+1/2 ln w
    i
    - ln p
    i

    d
    j
    (x) for all i j

    d
    i
    =(x- m
    i
    )
    t
    w
    i
    -1
    (x- m
    i
    )

    You might also like

    pFad - Phonifier reborn

    Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

    Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


    Alternative Proxies:

    Alternative Proxy

    pFad Proxy

    pFad v3 Proxy

    pFad v4 Proxy