Methods and apparatus to detect the location of sound sources external to computing devices

Patent metadata

Field	Value
Publication number	US-11860288-B2
Application number	US-202016914064-A
Country	US
Kind code	B2
Filing date	Jun 26, 2020
Priority date	Jun 26, 2020
Publication date	Jan 2, 2024
Grant date	Jan 2, 2024

Abstract

Methods, apparatus, systems, and articles of manufacture to detect the location of sound sources external to computing devices are disclosed. An apparatus, to determine a direction of a source of a sound relative to a computing device, includes a cross-correlation analyzer to generate a vector of values corresponding to a cross-correlation of first and second audio signals corresponding to the sound. The first audio signal is received from a first microphone of the computing device. The second audio signal is received from a second microphone of the computing device. The apparatus also includes a location analyzer to use a machine learning model and a set of the values of the vector to determine the direction of the source of the sound.

First claim

What is claimed is:

1. An apparatus to determine a direction of a source of a sound relative to a computing device, the apparatus comprising:
a cross-correlation analyzer to generate a vector of values corresponding to a cross-correlation of first and second audio signals corresponding to the sound, the first audio signal received from a first microphone of the computing device, the second audio signal received from a second microphone of the computing device; and
a location analyzer to use a machine learning model and a set of the values of the vector to determine the direction of the source of the sound, the set of the values corresponding to a segment of the vector including at least a threshold number of the values of the vector, the threshold number being at least twice a number of samples corresponding to a time difference of arrival for a sound originating at a point collinear with the first and second microphones.

2. The apparatus of claim 1 , wherein the location analyzer is to use the machine learning model to determine the direction of the source of the sound across 360 degrees of space surrounding the computing device without feedback from additional microphones other than the first and second microphones.

3. The apparatus of claim 1 , wherein the machine learning model is to distinguish between a first source of sound located in front of the computing device and a second source of sound located behind the computing device.

4. The apparatus of claim 1 , wherein the first and second microphones are to be spaced apart on a surface of the computing device, the first and second microphones facing a same direction as the surface.

5. The apparatus of claim 4 , further including using the machine learning model to determine the direction of the source of the sound from among a plurality of possible directions, the plurality of possible directions including first directions distributed across a front 180 degree region and second directions distributed across a back 180 degree region, the front 180 degree region corresponding to an area in front of the surface, the back 180 degree region corresponding to an area behind the surface.

6. The apparatus of claim 1 , wherein the vector corresponds to a generalized cross correlation with phase transform of the first and second audio signals.

7. The apparatus of claim 1 , wherein the set of the values corresponds to less than all the values in the vector.

8. The apparatus of claim 1 , wherein the segment of the vector includes more than 1% of all of the values.

9. The apparatus of claim 1 , wherein the set of the values corresponds to a mid-section of the vector, the mid-section excluding ones of the values on either side of the mid-section.

10. The apparatus of claim 1 , wherein the segment of the vector surrounds a peak value in the vector.

11. The apparatus of claim 1 , further including a response generator to generate a response based on the direction of the source of the sound.

12. The apparatus of claim 11 , wherein the response generator is to at least one of:
isolate the sound when the direction of the source of the sound is within a threshold angle of a first direction; or
reduce noise associated with the sound when the direction of the source of the sound is outside of the threshold angle of the first direction.

13. The apparatus of claim 11 , wherein the response generator is to:
in response to the direction of the source corresponding to a front of the computing device, identify the sound as originating from a user of the computing device; and
in response to the direction of the source corresponding to a rear of the computing device, disregard the sound.

14. The apparatus of claim 1 , wherein the machine learning model is implemented by a shallow neural network.

15. A non-transitory computer readable medium comprising instructions that, when executed, cause a machine to at least:
generate a vector of values corresponding to a cross-correlation of first and second audio signals corresponding to a sound, the first audio signal received from a first microphone of a computing device, the second audio signal received from a second microphone of the computing device; and
using a machine learning model and a set of the values of the vector to determine a direction of a source of the sound, the set of the values of the vector corresponds to a segment of the vector including at least a threshold number of the values, the threshold number being at least twice a number of samples corresponding to a time difference of arrival for a sound originating at a point collinear with the first and second microphones.

16. The computer readable medium of claim 15 , wherein the instructions further cause the machine to use the machine learning model to determine the direction of the source of the sound across 360 degrees of space surrounding the computing device without feedback from additional microphones other than the first and second microphones.

17. The computer readable medium of claim 15 , wherein the set of the values corresponds to less than all the values in the vector.

18. The computer readable medium of claim 15 , wherein the set of the values corresponds to a mid-section of the vector, the mid-section excluding ones of the values on either side of the mid-section.

19. The computer readable medium of claim 15 , wherein the instructions further cause the machine to generate a response based on the direction of the source of the sound.

20. A method to determine a direction of a source of a sound relative to a computing device, the method comprising:
generating a vector of values corresponding to a cross-correlation of first and second audio signals corresponding to the sound, the first audio signal received from a first microphone of the computing device, the second audio signal received from a second microphone of the computing device; and
using a machine learning model and a set of the values of the vector to determine the direction of the source of the sound, the set of the values of the vector corresponds to a segment of the vector including at least a threshold number of the values, the threshold number being at least twice a number of samples corresponding to a time difference of arrival for a sound originating at a point collinear with the first and second microphones.

21. The method of claim 20 , further including using the machine learning model to determine the direction of the source of the sound across 360 degrees of space surrounding the computing device without feedback from additional microphones other than the first and second microphones.

22. An apparatus to determine a direction of a source of a sound relative to a computing device, the apparatus comprising:
means for generating a vector of values corresponding to a cross-correlation of first and second audio signals corresponding to the sound, the first audio signal received from a first microphone of the computing device, the second audio signal received from a second microphone of the computing device; and
means for using a machine learning model and a set of the values of the vector to determine the direction of the source of the sound, the set of the values of the vector corresponds to a segment of the vector including at least a threshold number of the values, the threshold number being at least twice a number of samples corresponding to a time difference of arrival for a sound originating at a point collinear with the first and second microphones.

Assignees

• Intel Corp

Inventors

• Cordourier Maruri Hector
• Kupryjanow Adam
• Duzinkiewicz Karol
• Camacho Perez Jose Rodrigo
• Meyer Paulo Lopez
• Esquivel Julio Zamora
• Ibarra Von Borstel Alejandro
• Huang Jonathan

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G01S3/8083 Primary

  determining direction of source · CPC title

• G01S3/86

  with means for eliminating undesired waves, e.g. disturbing noises · CPC title

• G06N3/08

  Learning methods · CPC title