Method of and means for adaptively filtering near-stationary noise from speech

• US 4,025,721 A
• Filed: 05/04/1976
• Issued: 05/24/1977
• Est. Priority Date: 05/04/1976
• Status: Expired due to Term

• Current Assignee: Intellitech Coproration
• Original Assignee: Biocommunications Research Corporation
• Inventors: Graupe, Daniel, Causey, G. Donald
• Primary Examiner(s): Cooper, William C.
• Assistant Examiner(s): Kemeny, E. S.
• Application Number:
  US05/683,234
• Time in Patent Office: 385 Days
• Field of Search: 179/1 P, 179/1 D, 179/1 SA, 179/107 R, 179/107 FD
• US Class Current: 704/227
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: H03G 5/18 in untuned amplifiers
• CPC Class Codes: H04R 25/453 electronically
• CPC Class Codes: H04R 25/505 using digital signal processing

Abstract

By identifying and analyzing the properties of the parameters of an input signal that contains speech in the presence of simultaneously occuring near-stationary noise, pauses between speech intervals as well as the termination of such noise can be recognized. When a pause interval containing noise is recognized, the parameters identified during such interval are used to set the parameters of an adaptive filter through which the input signal is passed during subsequent intervals of speech and until the noise terminates. During the time the input signal passes through the filter, the near-stationary noise is filtered out. In response to recognition of the termination of noise, the input signal is caused to by-pass the filter which is then prepared to accept the parameters of noise occuring in a subsequent pause.

Claims

1. A method for adaptively filtering near stationary, long-duration noise from speech comprising the steps of:

(a) periodically identifying the parameters representing the characteristics of an input signal comprising said noise and speech;

(b) detecting the occurrence of pauses containing said noise between speech intervals;

(c) adjusting the parameters of a filter to those identified in response to detection of a pause containing said noise; and

(d) passing the input signal through the filter followng the detected noise containing pause only until cessation of said noise is detected.

2. A method according to claim 1 including the steps of:

(a) detecting the occurrence of pauses without said noise between speech intervals; and

(b) causing the input signal to bypass during this speech interval following the detected pause without noise.

3. A method according to claim 1 including the step of causing the input signal to bypass the filter in response to termination of said noise within a pause between speech intervals.

4. A method according to claim 3 including the step of causing the input signal to bypass the filter in response to termination of said noise during a speech interval.

5. A method according to claim 1 wherein the parameter identification is a time-series identification.

6. A method according to claim 5 wherein the parameter identification is an auto-regressive model identification.

7. A method according to claim 5 wherein the parameter identification is an auto-regressive moving average model identification.

8. A method according to claim 1 wherein the parameter identification is spectral identification and the frequency domain obtained by taking a transform of the input signal.

9. A method according to claim 1 wherein the parameter identification is obtained by spectral identification of a time-series model based on taking its Fourier transform.

10. A method according to claim 1 wherein the parameter identification is obtained by spectral identification of a time-series model based on transforming the model to a frequency domain model.

11. A method according to claim 1 wherein the parameter identification uses an auto-regressive model transformed into a frequency domain model.

12. An adaptive filter for filtering near stationary, long-duration noise from speech comprising:

(a) discriminator means for periodically identifying the parameter representing the characteristics of an input signal comprising said noise and speech, and for detecting pauses containing said noise between speech intervals;

(b) filter means having adjustable parameters;

(c) control means responsive to detection of the pause containing noise for adjusting the parameters of the filter means to the identified parameters and for passing the input signal through the filter means following the detected noise-containing pause only until cessation of said noise.

13. An adaptive filter according to claim 12 wherein the discriminator means detects pauses without noise and said control means is responsive to detection of a pause without noise for causing the input signal to bypass the filter means.

14. An adaptive filter according to claim 12 including means responsive to termination of noise during a pause for causing the input signal to bypass the filter, and for preparing the filter means to accept a new set of parameters.

15. An adaptive filter according to claim 12 wherein the discriminator means includes means for computing the sampled variance of the input signal, and the adaptive filter includes means responsive to a sudden drop in the sampled variance for causing the input signal to by-pass the filter means.

16. An adaptive filter according to claim 12 wherein the filter means is in the form of an augmented Kalman filter, namely a colored measurement noise Kalman filter where certain noise vector elements augment the state vector.

17. An adaptive filter according to claim 12 wherein the filter means is in the form of an adjustable notch filter, namely a notch filter wherein the frequency location of the notch is adjustable.

18. An adaptive filter according to claim 12 wherein the filter means is in the form of a Wiener filter.

19. An adaptive filter according to claim 12 wherein the filter is a linear least square filter.

20. An adaptive filter according to claim 12 wherein the filter means is in the form of an augmented Kalman filter for colored measurement noise.

21. An adaptive filter according to claim 12 wherein the filter means is in the form of an augmented Kalman filter and the adaptive filter includes means for computing the state error covariance matrix P of the Kalman filter, and means responsive to a threshold increase in the value of the P matrix for causing the input signal to by-pass the filter means.

22. An adaptive filter according to claim 12 wherein the filter means is in the form of an augmented Kalman filter, and the adaptive filter includes means for computing the covariance matrix P of the Kalman filter, means for computing the sampled variance of the input signal, and means responsive to a sudden drop in the sampled variance simultaneously with a threshold increase in the value of the P matrix for causing the input signal to by-pass the filter means.

23. An adaptive filter according to claim 15 including means for monitoring the filter output and responsive to maintain one of the output when the sampled variance decreases suddenly for causing the input signal to by-pass the filter means.

24. A voice communication system including the adaptive filter of claim 12 in combination with an amplifier driven speaker whose input is derived from the output of the filter means during occurrence of noise of the type described.

25. A hearing aid system comprising an adaptive filter according to claim 12 in combination with a hearing aid whose input is the output of said filter means.

26. A hearing aid system according to claim 24 wherein the adaptive filter is physically incorporated into the hearing aid.

27. A hearing aid system according to claim 24 wherein the adaptive filter is physically separate from the hearing aid but is linked thereto by a wire.

28. A hearing aid system according to claim 24 wherein the adaptive filter is physically separate from the hearing aid but is linked thereto by a radio link.