Method and system of acoustic angle of arrival detection

Patent metadata

Field	Value
Publication number	US-11435429-B2
Application number	US-201916359101-A
Country	US
Kind code	B2
Filing date	Mar 20, 2019
Priority date	Mar 20, 2019
Publication date	Sep 6, 2022
Grant date	Sep 6, 2022

Abstract

A system, article, and method of acoustic angle of arrival detection uses both same-time and delayed-time audio signal value comparisons in a time domain that are input to a classifier neural network.

First claim

What is claimed is:

1. A computer-implemented method of acoustic angle of arrival detection comprising:
receiving, by at least one processor, multiple audio signals from at least two microphone channels, wherein each channel provides an audio signal from a different microphone;
sampling the audio signals at a plurality of time frames to form audio samples of each channel;
generating same-time comparison values comprising comparing the audio samples of different channels and generated at a same time frame;
generating delay comparison values comprising comparing the audio samples that are both of different channels and of at least partially different time frames; and
determining an acoustic angle of arrival of acoustics forming the audio signals and comprising using both the same-time comparison values and the delay comparison values.

2. The method of claim 1 wherein the comparing related to both the same-time and delay comparisons comprises determining the root mean square of a difference between audio samples that are being compared.

3. The method of claim 1 comprising inputting a version of the same-time and delay comparison values into a classifier neural network to form the acoustic angle of arrival.

4. The method of claim 3 wherein each output of the neural network is associated with an angle of arrival relative to a reference direction.

5. The method of claim 3 wherein the same-time and delay comparison values together form a concatenated vector that is input to the neural network.

6. The method of claim 3 wherein the neural network only has three layers: an input layer, one hidden layer, and an output layer.

7. The method of claim 1 comprising normalizing the audio samples before performing the comparisons.

8. The method of claim 1 comprising obtaining the angle of arrival without converting the audio values into a frequency domain.

9. The method of claim 1 comprising obtaining the angle of arrival without using specific function hardware directed to obtaining the angle of arrival from the audio signals.

10. The method of claim 1 comprising using a delay measured by the number of the samples of time frames and being two to form the delay comparison values.

11. The method of claim 1 wherein the audio sample of each channel is compared to audio samples of all other channels to form both the same-time and delay comparison values.

12. A system of acoustic angle of arrival detection, comprising:
at least two microphones to receive at least two acoustic signals in an actual acoustic environment;
at least one memory communicatively coupled to the at least two microphones; and
at least one processor communicatively connected to the at least two microphones and at least one memory, the at least one processor being arranged to operate by:
receiving multiple audio signals from at least two microphone channels, wherein each channel provides an audio signal from one of the microphones;
sampling the audio signals at a plurality of times to form audio values of each channel;
generating comparison values comprising differencing the audio values of different channels and determining a root mean square average of the differences; and
determining an acoustic angle of arrival of acoustics forming the audio signals and comprising using the comparison values.

13. The system of claim 12 wherein the comparison values are input into a classifier neural network having outputs that individually indicate an angle of arrival relative to a reference direction.

14. The system of claim 12 wherein generating comparison values comprises:
generating same-time comparison values comprising comparing the audio values generated at a same time frame.

15. The system of claim 14 wherein the audio value of each channel is compared to the audio value of all other channels and of the same time frame to form the same-time comparison values.

16. The system of claim 14 wherein generating comparison values comprises:
generating delay comparison values comprising comparing the audio values of at least partially different time frames.

17. The system of claim 16 wherein the audio values of multiple channels at one time frame is compared to the audio values of other channels at a time frame that is different than the one time frame to form the delay comparison values.

18. The system of claim 16 wherein the at least one processor is arranged to operate by:
placing a version of the same-time and delay comparison values into a concatenated vector; and inputting the vector into a classifier neural network that has outputs associated with angles of arrival.

19. The system of claim 16 wherein the delay comparison values are formed by counting a delay measured by the number of the samples over time and being two to form the delay comparison values.

20. The system of claim 12 wherein the at least one processor is to operate by obtaining the angle of arrival from the audio values without converting the audio values to a frequency domain.

21. The system of claim 12 wherein the angle of arrival is obtained by using the audio signals without the use of a digital signal processor (DSP).

22. At least one non-transitory computer readable medium comprising a plurality of instructions that in response to being executed on a computing device, causes the computing device to operate by:
receiving same-time comparison values generated by comparing audio values of a plurality of channels and generated at a same time frame;
receiving delay comparison values comprising comparing audio values of at least partially different time frames; and
inputting the same-time and delay comparison values to a classifier neural network that outputs probabilities of an angle of arrival being associated with one or more angles.

23. The medium of claim 22 wherein the same-time comparison values and delay comparison values are concatenated into a single vector before being input to the classifier neural network.

24. The medium of claim 23 wherein the comparisons are performed by computing the root mean square of a difference between audio values being compared, and without converting the audio values into a frequency domain.

25. The medium of claim 22 wherein the different time frames are a predetermined number of samples apart.

Assignees

• Intel Corp

Inventors

• Cordourier Maruri Hector
• Camacho Perez Jose
• Lopez Meyer Paulo
• Zamora Esquivel Julio
• Huang Jonathan

Classifications

• G01S3/8083

  determining direction of source · CPC title

• H04R2420/07

  Applications of wireless loudspeakers or wireless microphones · CPC title

• H04R2430/20

  Processing of the output signals of the acoustic transducers of an array for obtaining a desired directivity characteristic (H04R2203/12 takes precedence) · CPC title

• G01S3/8034 Primary

  wherein the signals are derived simultaneously · CPC title

• H04R2201/401

  2D or 3D arrays of transducers · CPC title