NEURAL NETWORK-BASED LOUDSPEAKER MODELING WITH A DECONVOLUTION FILTER

• US 20170303039A1
• Filed: 04/14/2016
• Published: 10/19/2017
• Est. Priority Date: 04/14/2016
• Status: Active Grant

• Current Assignee: Harman International Industries Incorporated (Samsung Electronics Co. Ltd.)
• Original Assignee: Harman International Industries Incorporated (Samsung Electronics Co. Ltd.)
• Inventors: Iyer, Ajay, Button, Douglas J.
• Granted Patent: US 10,425,730 B2
• Time in Patent Office: Days
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G06F 3/165 Management of the audio stream, e.g. setting of volume, audio stream path
• CPC Class Codes: H04R 29/001 for loudspeakers H04R29/007 takes precedence
• CPC Class Codes: H04R 3/00 Circuits for transducers , loudspeakers or microphones
• CPC Class Codes: H04R 3/04 for correcting frequency response
• CPC Class Codes: H04S 7/305 Electronic adaptation of stereophonic audio signals to reverberation of the listening space H04S7/301 takes precedence

Abstract

A technique for controlling a loudspeaker system with an artificial neural network includes filtering, with a deconvolution filter, a measured system response of a loudspeaker and a reverberant environment in which the loudspeaker is disposed to generate a filtered response, wherein the measured system response corresponds to an audio input signal applied to the loudspeaker while the loudspeaker is disposed in the reverberant environment. The techniques further include generating, via a neural network model, an initial neural network output based on the audio input signal, comparing the initial neural network output to the filtered response to determine an error value, and generating, via the neural network model, an updated neural network output based on the audio input signal and the error value.

Claims

1. A computer-implemented method of generating a filter for a loudspeaker control system, the method comprising:

receiving, via a sensor, a measured system output of a loudspeaker and a reverberant environment in which the loudspeaker is disposed;

extracting a head portion and tail portion from the measured system output, wherein the head portion includes a measured output of the loudspeaker and the tail portion includes a response of the reverberant environment;

determining an inverse of the response of the reverberant environment; and

generating the filter based on the inverse of the response.

2. The method of claim 1, further comprising generating a system output related to the loudspeaker and the reverberant environment in which the loudspeaker is disposed by applying an input stimulus to the loudspeaker.

3. The method of claim 1, wherein the head portion does not include any portion of the response of the reverberant environment.

4. The method of claim 1, wherein the head portion does not include a first reflection of an output of the loudspeaker.

5. The method of claim 1, wherein the measured system output includes at least a first reflection of an output of the loudspeaker and at least a portion of the output of the loudspeaker.

6. The method of claim 1, wherein determining the inverse of the response of the reverberant environment comprises determining the inverse of the response of the reverberant environment based on the tail portion.

7. The method of claim 6, wherein determining the inverse of the response of the reverberant environment based on the tail portion comprises computing an impulse response function for the reverberant environment based on the tail portion, transforming the impulse response function for the reverberant environment into a first spectrum in the frequency domain, inverting the first spectrum to create an inverted spectrum, and transforming the inverted spectrum to the time domain.

8. The method of claim 1, wherein determining the inverse of the response of the reverberant environment comprises computing a first impulse response function for the loudspeaker based on the head portion, and computing a second impulse response function for the loudspeaker while disposed in the reverberant environment based on the measured system output.

9. The method of claim 8, wherein determining the inverse of the response of the reverberant environment further comprises transforming the first impulse response function into a first spectrum in the frequency domain, and transforming the second impulse response function into a second spectrum in the frequency domain.

10. The method of claim 9, wherein determining the inverse of the response of the reverberant environment further comprises dividing the first spectrum by the second spectrum to obtain a result, and transforming the result into the time domain .

11. The method of claim I, wherein the measured system output comprises an acoustic signal having a final amplitude that is approximately 60 db less than an initial amplitude.

12. A non-transitory computer-readable storage medium including instructions that, when executed by one or more processors, configure the one or more processors to control a loudspeaker system with an artificial neural network, by performing the steps of:

filtering, with a deconvolution filter, a measured system response of a loudspeaker and a reverberant environment in which the loudspeaker is disposed to generate a filtered response, wherein the measured system response corresponds to an audio input signal applied to the loudspeaker while the loudspeaker is disposed in the reverberant environment;

generating, via a neural network model, an initial neural network output based on the audio input signal;

comparing the initial neural network output to the filtered response to determine an error value; and

generating, via the neural network model, an updated neural network output based on the audio input signal and the error value.

13. The system of claim 12, wherein the deconvolution filter is based on an inverse of a response of the reverberant environment.

16. The system of claim 12, wherein the initial neural network output is configured to approximate the filtered response.

19. The method of claim 16, wherein the measured system response comprises at least one reverberation of an output of the loudspeaker.

20. The method of claim 16, wherein the deconvolution filter is based on an inverse of a response of the reverberant environment.

14. The system of claim 14, wherein the response of the reverberant environment includes a response to an output of the loudspeaker while the loudspeaker is disposed in the reverberant environment.

15. The system of claim 14, wherein the inverse of the response is determined based on a head portion of a system response of the loudspeaker when disposed in the reverberant environment and on the system response of the loudspeaker when disposed in the reverberant environment.

17. A system, comprising:

a loudspeaker disposed in a reverberant environment;

a memory storing a loudspeaker control algorithm; and

one or more processors that are coupled to the memory and, when executing the loudspeaker control algorithm, are configured to;

filter, with a deconvolution filter, a measured system response of the loudspeaker and the reverberant environment to generate a filtered response, wherein the measured system response corresponds to an audio input signal applied to the loudspeaker while the loudspeaker is disposed in the reverberant environment;

generate, via a neural network model, an initial neural network output based on the audio input signal;

compare the initial neural network output to the filtered response to determine an error value; and

generate, via the neural network model, an updated neural network output based on the audio input signal and the error value.

18. The system of claim 17, further comprising a sensor that is disposed in the reverberant environment and is configured to generate the measured system response.