Multi-channel adaptive speech signal processing system with noise reduction

• US 20060222184A1
• Filed: 09/23/2005
• Published: 10/05/2006
• Est. Priority Date: 09/23/2004
• Status: Active Grant

• Current Assignee: Cerence Operating Company (Cerence Inc.)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Buck, Markus, Haulick, Tim, Hetherington, Phillip A., Zakarauskas, Pierre
• Granted Patent: US 8,194,872 B2
• Time in Patent Office: Days
• US Class Current: 381/71.100
• CPC Class Codes: G10L 2021/02166 Microphone arrays; Beamforming
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: G10L 21/0216 characterised by the method used for estimating noise
• CPC Class Codes: H04R 2430/25 Array processing for suppression of unwanted side-lobes in directivity characteristics, e.g. a blocking matrix
• CPC Class Codes: H04R 2499/13 Acoustic transducers and sound field adaptation in vehicles
• CPC Class Codes: H04R 29/006 Microphone matching
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407

Abstract

An adaptive signal processing system eliminates noise from input signals while retaining desired signal content, such as speech. The resulting low noise output signal delivers improved clarity and intelligibility. The low noise output signal also improves the performance of subsequent signal processing systems, including speech recognition systems. An adaptive beamformer in the signal processing system consistently updates beamforming signal weights in response to changing microphone signal conditions. The adaptive weights emphasize the contribution of high energy microphone signals to the beamformed output signal. In addition, adaptive noise cancellation logic removes residual noise from the beamformed output signal based on a noise estimate derived from the microphone input signals.

Claims

1. A noise reduction signal processing system comprising:

multiple microphone signal inputs;

time delay compensation logic coupled to the microphone signal inputs and comprising time delay compensated microphone signal outputs;

noise reference logic coupled to the time delay compensation logic and comprising noise reference signal outputs;

an adaptive beamformer coupled to the time delay compensation logic and comprising a beamformed signal output, the adaptive beamformer generating a beamformed signal on the beamformed signal output; and

adaptive noise cancellation logic coupled to the noise reference signal outputs and the beamformed signal output and operable to generate a noise reference signal for removing noise from the beamformed signal.

2. The noise reduction signal processing system of claim 1, where the adaptive beamformer applies an adaptive real-valued weight to time delay compensated microphone signals provided on the time delay compensated microphone signal outputs.

3. The noise reduction signal processing system of claim 1, where the adaptive beamformer generates the beamformed signal according to:

4. The noise reduction signal processing system of claim 3, where ‘

A_m(n)’

is a repeatedly recalculated weight which adapts the adaptive beamformer over time to temporal changes in at least one of noise power and speech amplitude.

5. The noise reduction signal processing system of claim 3, where the repeatedly recalculated weight is a normalized repeatedly recalculated weight.

6. The noise reduction signal processing system of claim 1, where the noise reference logic comprises a blocking matrix.

7. The noise reduction signal processing system of claim 1, further comprising adaptive self-calibration logic coupled to the time delay compensation logic, the adaptive self-calibration logic operable to match at least one of phase and amplitude of time delay compensated microphone signals provided on the time delay compensated microphone signal outputs.

8. The noise reduction signal processing system of claim 1, further comprising adaptation control logic coupled to at least one of the adaptive beamformer and the adaptive noise cancellation logic.

9. The noise reduction signal processing system of claim 8, where the adaptation control logic initiates adaptation depending on at least one of:

instantaneous SNR, speech signal detection, speech signal energy level, and acoustic signal direction.

10. The noise reduction signal processing system of claim 1, where the multiple microphone signal inputs comprise a first directional microphone signal input and a second directional microphone signal input from microphones pointing in different directions.

11. The noise reduction signal processing system of claim 1, where the multiple microphone signal input comprise first sub-array microphone signal inputs and second sub-array microphone signal inputs from different microphone sub-arrays.

12. A method for reducing noise comprising:

receiving multiple microphone input signals;

applying a time delay compensation to the microphone input signals, thereby generating time delay compensated microphone output signals;

generating noise reference output signals based on the time delay compensated microphone output signals;

repeatedly updating weights in an adaptive beamformer responsive to temporal changes in the microphone input signals;

beamforming the time delay compensated microphone output signals into a beamformed signal based on the weights;

generating a noise reference signal based on the noise reference output signal; and

subtracting the noise reference signal from the beamformed signal.

13. The method of claim 12, where repeatedly updating comprises:

repeatedly updating real-valued weights.

14. The method of claim 12, where beamforming comprises determining a beamformed signal according to:

15. The method of claim 12, further comprising normalizing the weights.

16. The method of claim 12, where generating noise reference signals comprises:

generating noise reference signals using a blocking matrix.

17. The method of claim 12, further comprising applying adaptation control over updating the weights.

18. The method of claim 12, where generating an adaptive noise cancellation output comprises applying adaptation control over adaptive noise cancellation logic.

19. The method of claim 12, where receiving comprises:

receiving a first directional microphone input signal and a second directional microphone signal from microphone pointing in different directions.

20. The method of claim 12, where receiving comprises:

receiving a first microphone input signal and a second microphone input signal from different microphone sub-arrays.

21. A product comprising:

a machine readable medium; and

instructions encoded on the medium which cause a processor in a noise reduction signal processing system to perform a method comprising;

receiving multiple microphone input signals;

applying a time delay compensation to the microphone input signals, thereby generating time delay compensated microphone output signals;

generating noise reference output signals based on the time delay compensated microphone output signals;

repeatedly updating weights in an adaptive beamformer;

beamforming the time delay compensated microphone output signals into a beamformed signal based on the weights;

generating a noise reference signal based on the noise reference signal outputs; and

subtracting the noise reference signal from the beamformed signal.

22. The product of claim 21, where repeatedly updating comprises:

repeatedly updating real-valued weights.

23. The product of claim 21, where beamforming comprises determining the beamformed signal according to:

24. The product of claim 21, further comprising applying adaptation control over updating the weights.

25. The product of claim 21, where generating an adaptive noise cancellation output comprises applying adaptation control over adaptive noise cancellation logic.

26. The product of claim 21, where receiving comprises:

receiving a first directional microphone input signal and a second directional microphone signal from microphone pointing in different directions.

27. The product of claim 21, where receiving comprises:

receiving a first microphone input signal and a second microphone input signal from different microphone sub-arrays.

28. A noise reduction signal processing system comprising:

multiple microphone signal inputs comprising first directional microphone signal inputs and second directional microphone signal inputs from microphones pointing in different directions;

time delay compensation logic coupled to the microphone signal inputs and comprising time delay compensated microphone signal outputs;

adaptive self-calibration logic coupled to the time delay compensation logic, the adaptive self-calibration logic operable to match at least one of phase and amplitude of time delay compensated microphone output signals on the time delay compensated microphone signal outputs;

an adaptive blocking matrix coupled to the adaptive self-calibration logic and comprising noise reference signal outputs;

an adaptive beamformer coupled to the adaptive self-calibration logic which determines a beamformed signal according to;

$Y_w<br><br>\left(n,k\right)=\underset{m=1}{\overset{M}{\sum <br><br>}}<br><br>A_m<br><br>\left(n\right)<br><br>X_{T,m}<br><br>\left(n,k\right)$

where ‘

Y_w’

is the beamformed signal, ‘

n’

is a frequency bin index, ‘

k’

is a time index, there are ‘

M’

time delay compensated microphone output signals, ‘

A_m(n)’

is a repeatedly adapted real-valued weight, and ‘

X_T,m’

is a time delay compensated microphone output signal;

adaptive noise cancellation logic coupled to the noise reference signal outputs and the beamformed signal output and comprising an adaptive noise cancellation output, the adaptive noise cancellation logic operable to generate a noise reference signal on the adaptive noise cancellation output; and

summing logic for removing noise in the beamformed signal using the noise reference signal.

29. The noise reduction signal processing system of claim 28, where the adaptation control logic initiates adaptation of the adaptive beamformer when speech signal energy exceeds background noise by more than a threshold.

30. The noise reduction signal processing system of claim 28, where the adaptation control logic is also coupled to the adaptive noise cancellation logic, and where the adaptation control logic initiates adaptation of the adaptive noise cancellation logic in the substantial absence of speech signal energy and when noise is present.

31. The noise reduction signal processing system of claim 28, further comprising adaptation control logic coupled to the adaptive beamformer and the adaptive blocking matrix, the adaptation control logic operable to adapt the adaptive blocking matrix in response to adaptation of the adaptive beamformer.

28-1. (canceled)