Automatic speaker verification by non-linear time alignment of acoustic parameters

Claim 2

In an auditory verification system in which acoustic parameters of a test sample of an individual''s speech are matched for identity to like parameters of a reference sample of his speech, that improvement which includes the steps of: time adjusting said test sample parameters with said reference parameters according to a nonlinear registration schedule, measuring internal dissimilarities and irregularities between said time adjusted parameters and said average reference parameters, and verifying said individual''s identity on the basis of said measures of dissimilarities and irregularities.

Claim 3

In an auditory verification system in which acoustic parameters of a test sample of an individual''s speech are matched for identity to corresponding parameters of a reference sample of his speech, that improvement which comprises the steps of: preparing said reference sample of an individual''s speech from sets of parameters developed from a plurality of different utterances of a test phrase by said individual which have been mutually registered in time, and from a plurality of measures of variation between said different utterances, measuring internal dissimilarities and irregularities between said test sample parameters and said reference parameters, and verifying said individual''s identity on the basis of said measures of dissimilarities and irregularities.

Claim 4

In an auditory verification system in which acoustic parameters of a test sample of an individual''s speech are matched for identity to corresponding parameters of a reference sample of his speech, that improvement which comprises the steps of: preparing said reference sample of an individual''s speech from sets of parameters developed from a plurality of different utterances of a test phrase by said individual which have been mutually registered in time, and from a plurality of measures of variation between said different utterances, measuring internal dissimilarities and irregularities between said test sample parameters and said reference parameters, and verifying said individual''s identity on the basis of said measures of dissimilarities and irregularities.

Claim 5

In an auditory verification system in which acoustic parameters of a test sample of an individual''s speech are matched for identity to like parameters of a reference sample of his speech, that improvement which comprises the steps of: developing said reference sample from time registered values of a plurality of different speech signal parameters, developing a like plurality of different speech signal parameters from said test speech sample, time adjusting said test sample parameters with said reference parameters according to a nonlinear registration schedule, measuring internal dissimilarities and irregularities between said time adjusted parameters and said average reference parameters, and verifying said individual''s identity on the basis of said measures of dissimilarities and irregularities.

Claim 6

In an auditory verification system in which acoustic parameters of a test sample of an individual''s speech are matched for identity to like parameters of a reference sample of his speech, that improvement which comprises the steps of: developing said reference sample from time registered values of a plurality of different speech signal parameters, developing a like plurality of different speech signal parameters from said test speech sample, time adjusting said test sample parameters with said reference parameters according to a nonlinear registration schedule, measuring internal dissimilarities and irregularities between said time adjusted parameters and said average reference parameters, and verifying said individual''s identity on the basis of said measures of dissimilarities and irregularities.

Claim 7

In a speech signal verification system wherein selected speech signal parameters derived from a test phrase spoken by an individual to produce a sample are compared to reference parameters derived from the same test phrase spoken by the same individual, and wherein verification or rejection of the identity of the individual is determined by the similarities of said sample and reference parameters, means for bringing the time span of said sample parameters into temporal registration with the time span of said reference parameters, and means for temporally adjusting the time distribution of parameters of said sample withIn said adjusted time span to maximize similarities between said sample parameters and said reference parameters.

Claim 8

In a speech signal verification system wherein selected speech signal parameters derived from a test phrase spoken by an individual to produce a sample are compared to reference parameters derived from the same test phrase spoken by the same individual, and wherein verification or rejection of the identity of the individual is determined by the similarities of said sample and reference parameters, means for bringing the time span of said sample parameters into temporal registration with the time span of said reference parameters, and means for temporally adjusting the time distribution of parameters of said sample withIn said adjusted time span to maximize similarities between said sample parameters and said reference parameters.

Claim 9

The speech signal verification system as defined in claim 4, wherein said similarities between said sample parameters and said reference parameters are measured by the coefficient of correlation therebetween.

Claim 10

The speech signal verification system as defined in claim 5, wherein said temporal adjustment of parameters within said adjusted time span comprises, means for iteratively incrementing the time locations of selected parameter features until said measure of correlation between said sample parameters and said reference parameters does not increase significantly for a selected number of consecutive iterations.

Claim 11

The speech signal verification system as defined in claim 4, wherein said means for temporally adjusting said parameters within said adjusted time span comprises, means for temporally transforming said sample parameters, designated s(t), into a set of parameters, designated s( Tau ), in which Tau (t) a+bt+q(t), in which a and b are constants selected to align the end points of said time span of said sample parameters with the end points of said reference parameters, and in which q(t) is a nonlinear function which defines the distribution of parameter values within said time span.

Claim 12

The speech signal verification system as defined in claim 7, wherein said nonlinear function q(t) is a continuous piece-wise linear function described by N selected amplitude values qi and time values ti within said time span, wherein i 0, 1, . . .N.

Claim 13

An auditory speech signal verification system, which comprises, in combination, means for analyzing a plurality of individual utterances of a test phrase spoken by an individual to develope a prescribed set of acoustic parameter signals for each utterance, means for developing from each of said sets of parameter signals a reference set of parameter signals and a set of signals which denotes variations between parameter signals used to develop said reference set of signals, means for storing a set of reference parameter signals and a set of variation signals for each of a number of different individuals, means for analyzing a sample utterance of said test phrase spoken by an individual purported to be one of said number of different individuals to develop a set of acoustic parameter signals, means for adjusting selected parameter signals of said sample to bring the time scale of said utterance represented by said parameters into registry with the time scale of a designated one of said stored reference utterances represented by said reference parameters, said means including means for adjusting selected values of said sample parameter signals to maximize similarities between said sample utterance and said reference utterance, means responsive to said reference parameter signals, said adjusted sample parameter signals, and said variation signals for developing a plurality of signals representative of selected similarities between each of said sample parameters and each of said corresponding reference parameters, means for developing signals representative of the extent of adjustment employed to register said time scales, means responsive to said plurality of similarity signals and said signals representative of the extent of adjustment for developing a signal representative of the overall degree of similarity between said sample utterance and said designated reference utterance, and threshold comparison means supplied with said overall similarity signal for matching the magnitude of said similarity signal to the magnitude of a stored threshold signal, and for issuing an ''''accept'''' signal for similarity signals above threshold, a ''''reject'''' signal for signals below threshold, and a ''''no decision'''' signal for simiLarity signals within a prescribed narrow range of signal magnitudes near said threshold magnitude.

Claim 14

An auditory speech signal verification system, which comprises, in combination, means for analyzing a plurality of individual utterances of a test phrase spoken by an individual to develope a prescribed set of acoustic parameter signals for each utterance, means for developing from each of said sets of parameter signals a reference set of parameter signals and a set of signals which denotes variations between parameter signals used to develop said reference set of signals, means for storing a set of reference parameter signals and a set of variation signals for each of a number of different individuals, means for analyzing a sample utterance of said test phrase spoken by an individual purported to be one of said number of different individuals to develop a set of acoustic parameter signals, means for adjusting selected parameter signals of said sample to bring the time scale of said utterance represented by said parameters into registry with the time scale of a designated one of said stored reference utterances represented by said reference parameters, said means including means for adjusting selected values of said sample parameter signals to maximize similarities between said sample utterance and said reference utterance, means responsive to said reference parameter signals, said adjusted sample parameter signals, and said variation signals for developing a plurality of signals representative of selected similarities between each of said sample parameters and each of said corresponding reference parameters, means for developing signals representative of the extent of adjustment employed to register said time scales, means responsive to said plurality of similarity signals and said signals representative of the extent of adjustment for developing a signal representative of the overall degree of similarity between said sample utterance and said designated reference utterance, and threshold comparison means supplied with said overall similarity signal for matching the magnitude of said similarity signal to the magnitude of a stored threshold signal, and for issuing an ''''accept'''' signal for similarity signals above threshold, a ''''reject'''' signal for signals below threshold, and a ''''no decision'''' signal for simiLarity signals within a prescribed narrow range of signal magnitudes near said threshold magnitude.

Claim 15

An auditory speech signal verification system, as defined in claim 9, wherein, said means for developing a plurality of signals representative of selected similarities includes, means for measuring a plurality of different speech signal characteristics for similarity in each of a number of time subintervals within the interval of said designated time scale.

Claim 16

An auditory speech signal verification system, as defined in claim 10, wherein, said different speech signal characteristics are based, respectively, on (1) the difference in average values between said reference speech signal parameters and said sample speech signal parameters, (2) the squared difference in linear components between the two, (3) the squared difference in quadratic components between the two, (4) the correlation between the two in each of said subintervals, and (5) the correlation between the two over said entire interval of said designated time scale.

Claim 17

An auditory speech signal verification system, as defined in claim 11, wherein, each of said signals representative of selected similarities between each of said sample parameters and each of said corresponding reference parameters is scaled in accordance with the magnitudes of said variation signals.