Voice activity detection
First Claim
1. Voice activity detection apparatus comprising:
- (i) means for receiving an electrical input signal in which the presence or absence of signals representing speech is to be detected;
(ii) means responsive to said means for receiving for periodically adaptively generating an electrical signal representing an estimated noise signal component of the input signal by producing the autocorrelation coefficients Ai of the impulse response of a FIR filter having a response approximating the inverse of the short term spectrum of the noise signal component;
(iii) means responsive to said means for receiving for periodically forming from the input signal and the estimated noise representing signal an electrical signal representing a measure M of the spectral similarity between a portion of the input signal and the said estimated noise signal component, said measure forming means comprises means for producing electrical signals representing the autocorrelation coefficients Ri of the input signal, and means connected to receive Ri and Ai signals, and to calculate the measure M therefrom; and
(iv) electrical means responsive to said means for forming for comparing the electrical signals representing said measure with a threshold value representing signal to produce an electrical output indicating the presence or absence of speech in the electrical input signal.
1 Assignment
0 Petitions
Accused Products
Abstract
Voice activity detector (VAD) for use in an LPC coder in a mobile radio system uses autocorrelation coefficient R0, R1 . . . of the input signal, weighted and combined, to provide a measure M which depends on the power within that part of the spectrum containing no noise, which is thresholded against a variable threshold to provide a speech/no speech logic output. The measure is formula (I), where Hi are the autocorrelation coefficients of the impulse response of an Nth order FIR inverse noise filter derived from LPC analysis of previous non-speech signal frames. Threshold adaption and coefficient update are controlled by a second VAD response to rate of spectral change between frames.
167 Citations
23 Claims
-
1. Voice activity detection apparatus comprising:
-
(i) means for receiving an electrical input signal in which the presence or absence of signals representing speech is to be detected; (ii) means responsive to said means for receiving for periodically adaptively generating an electrical signal representing an estimated noise signal component of the input signal by producing the autocorrelation coefficients Ai of the impulse response of a FIR filter having a response approximating the inverse of the short term spectrum of the noise signal component; (iii) means responsive to said means for receiving for periodically forming from the input signal and the estimated noise representing signal an electrical signal representing a measure M of the spectral similarity between a portion of the input signal and the said estimated noise signal component, said measure forming means comprises means for producing electrical signals representing the autocorrelation coefficients Ri of the input signal, and means connected to receive Ri and Ai signals, and to calculate the measure M therefrom; and (iv) electrical means responsive to said means for forming for comparing the electrical signals representing said measure with a threshold value representing signal to produce an electrical output indicating the presence or absence of speech in the electrical input signal. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
6. Apparatus according to claim 1 or 4, in which ##EQU8##
-
7. Apparatus according to claims 1 or 4, in which said generating means comprises a buffer connected to store data from which the autocorrelation coefficients Ai of the said filter response may be obtained, in which the said filter response is periodically calculated from the signal by LPC analysis means, the apparatus being so connected and controlled that the measure M is calculated using the said stored data, and the said stored data is updated only from periods in which speech is indicated to be absent.
-
8. Apparatus according to claim 7 further comprising second voice activity detection means responsive to said input signal for indicating the absence of speech to control the updating of the stored data.
-
9. Apparatus according to claims 1 or 4, further comprising means for adjusting said threshold value during periods when speech is indicated to be absent.
-
10. Apparatus according to claim 9 further comprising second voice activity detection means responsive to said input signal to produce a control signal indicating the presence or absence of speech, said adjusting means being responsive to said control signal to prevent adjustment of said threshold value when speech is present.
-
11. Apparatus according to claim 9 in which said threshold value is, when adjusted, adjusted to be equal to the mean of the measure plus a term which is a fraction of the standard deviation of the measure.
-
12. Apparatus according to claim 10 further comprising means for adjusting the said threshold value during periods when speech is indicated to be absent, said second voice activity detection means serving also to prevent adjustment of the threshold value when speech is present.
-
13. Apparatus according to claim 10 in which said second voice activity detection means comprises means for generating a measure of the spectral similarity between a portion of the input signal and earlier portions of the input signal.
-
14. Apparatus according to claim 13 in which the similarity measure generating means of said second voice activity detection means comprises means for providing, from LPC filter data and autocorrelation data relating to a present portion of the input signal, a present distortion measure;
- means for providing an equivalent past frame distortion measure corresponding to a preceding portion of the input signal, and means for generating a signal indicating the degree of similarity therebetween as an indicator of speech presence or absence.
-
15. Apparatus according to claim 13, in which said second voice activity detection means further comprises voiced speech detection means comprising pitch analysis means, for generating a signal indicative of the presence of voiced speech, upon which the output of said second voice activity detection means also depends.
-
-
16. Voice activity apparatus comprising:
-
(i) means for receiving an electrical signal in which the presence or absence or signals representing speech is to be detected; (ii) means responsive to said means for receiving for periodically adaptively generating an electrical signal representing an estimated noise signal component of the input signal, said generating means including analysis means operable to produce electrical signals representative of the coefficients of a filter having a spectral response which is the inverse of the frequency spectrum of the estimated noise signal component; (iii) means responsive to said means for periodically adaptively generating for periodically forming from the input signal and the estimated noise representing signal and electrical signal representing a measure of a spectral similarity between a portion of the input signal and the said estimated noise signal component, the measure being proportional to a zero-order autocorrelation of the input signal after filtering by a filter having the said coefficients; and (iv) electrical means for comparing the measure with a threshold value to produce an output indicating the presence or absence of speech.
-
-
17. A method of detecting voice activity representing signals in an electrical input signal, comprising
(a) periodically adaptively generating an electrical signal representing an estimated noise signal component of the input signal, and producing signals representing the coefficients of a filter having a spectral response which is the inverse of the frequency spectrum of the estimated noise signal component; -
(b) periodically forming from the input signal and the estimated noise representing signal an electrical signal representing a measure of the spectral similarity between a portion of the input signal and the said estimated noise signal component, the measure being proportional to a zero-order autocorrelation of the input signal after filtering by a filter having the said coefficients; and (c) electrically comparing the measure with a threshold valve to produce an output indicating the presence or absence of speech.
-
-
18. Voice activity detection apparatus comprising:
-
(i) means for receiving an electrical input signal in which the presence or absence of signals representing speech is to be detected; (ii) analysis means responsive to said means for receiving operable to produce electrical signals representing the coefficients of a filter having a spectral response which is the inverse of the frequency spectrum of the input signal; (iii) means for periodically adaptively generating an electrical signal representing an estimated noise signal component of the input signal; (iv) electrical means responsive to said analysis means and said estimated noise generating means for periodically forming from the filter coefficients and the estimated noise representing signal further signals representing a measure of a spectral similarity between a portion of the input signal and the same estimated noise signal component, the measure being proportional to a zero-order autocorrelation of the noise representing signal after filtering by a filter having the same coefficients; and (v) means for comparing the measure with a threshold value to produce an output indicating the presence or absence of speech.
-
-
19. A method of detecting voice activity representing signals in an electrical input signal, comprising:
-
(a) producing electrical signals representing the coefficients of a filter having a spectral response which is the inverse of the frequency spectrum of the input signal; (b) periodically adaptively generating electrical signals representing an estimated noise signal component of the input signal; (c) periodically forming from the filter coefficients and the estimated noise representing signal an electrical signal representative of a measure of the spectral similarity between a portion of the input signal and the said estimated noise signal component, the measure being proportional to the zero-order autocorrelation of the noise representing signal after filtering by a filter having the said coefficients; and (d) comparing the measure with a threshold value to produce an output indicating the presence or absence of speech.
-
-
20. A voice activity detection apparatus comprising:
-
(i) a first voice activity detector which operates by forming electrical signals representing a measure of a spectral similarity between an electrical input signal and a speech free stored portion of an input signal to produce an electrical output signal indicating the presence or absence of speech in the input signal; (ii) a store for containing the stored portion of the input signal; and (iii) an auxiliary voice activity detector responsive to said electrical input signal to produce a second signal indicating the presence or absence of speech in the input signal, said second signal alone controlling the updating of said store, the auxiliary voice activity detector operating by forming an electrical signal representing a measure of a spectral similarity between a current input signal and an earlier portion of the input signal.
-
-
21. A voice activity detection apparatus comprising:
-
(i) means for receiving an electrical input signal in which the presence or absence of signals representing speech is to be detected; (ii) a store for storing an estimated noise representation signal; (iii) means responsive to said means for receiving for periodically forming from the input signal and the stored estimated noise representation signal an electrical signal representing a measurement of the spectral similarity between a portion of the input signal and the said estimated noise signal component; (iv) electrical means for comparing the measure with a threshold value to produce an output indicating the presence or absence of speech; (v) an auxiliary voice activity detector, operating by forming an electrical signal representing a measure of spectral similarlity between the input signal and a preceding portion of the input signal to produce a control signal indicating the presence or absence of speech; and (vi) store updating means operable to update the store from said electrical input signal only when said control signal indicates that speech is absent. - View Dependent Claims (22, 23)
-
Specification