Loudspeaker Neural Network Training with Reverberation Deconvolution

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for training artificial neural networks for loudspeaker control systems are impractical in reverberant environments due to the interference of acoustic reflections, which prevents accurate modeling of loudspeaker behavior.

Innovation Solution

A deconvolution filter is generated to reduce reverberant effects by at least 20 dB, allowing neural networks to be trained in reverberant environments and correct for linear and non-linear distortion in loudspeaker responses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If training is performed in an anechoic chamber to achieve accurate loudspeaker modeling, then measurement precision is improved, but device complexity and manufacturing scalability deteriorate

Engineering Contradiction:
Improveloudspeaker modeling accuracyVSAvoidanechoic chamber requirement
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

A deconvolution filter is introduced as an intermediary processing step between the microphone measurement and the neural network training input. The filter removes reverberant components from the measured impulse response, allowing accurate loudspeaker modeling to be performed using ordinary rooms instead of anechoic chambers.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of manufacture

If training is performed in a reverberant environment to improve scalability and ease of manufacture, then device complexity is reduced, but measurement precision deteriorates due to acoustic reflections

Engineering Contradiction:
Improvetraining environment accessibilityVSAvoidloudspeaker response accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent converts the harmful effect of room reverberations into a beneficial process. By deliberately measuring in reverberant environments and then applying deconvolution filtering to remove the reverberant components, the system achieves accurate loudspeaker modeling using ordinary rooms, thereby improving scalability while maintaining precision.

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

3Measurement precision

If deconvolution filtering is applied to remove reverberations, then measurement precision is improved, but processing time and computational complexity increase

Engineering Contradiction:
Improvereverberation reduction accuracyVSAvoidfilter generation and application time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The deconvolution filter is generated and applied during the offline training phase before neural network deployment. By performing the computationally intensive deconvolution operation in advance during system setup rather than in real-time during operation, the patent achieves high measurement precision while minimizing impact on operational time requirements.

Inventive Principle:
Principle #10Preliminary action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Enables accurate characterization and control of loudspeakers in reverberant environments by removing distortion caused by room reflections, improving the quality of loudspeaker output.

Implementation Method 1

a response of a reverberant environment in which a loudspeaker is disposed can be reduced by 20 db or more by a deconvolution filter

Methodology Applied
Scientific EffectDeconvolution:

Data Source

PatentEP3232686B1Neural network-based loudspeaker modeling with a deconvolution filter
Publication Date: 2021.03.03 HARMAN INT IND INC
  • EP3232686B1 patent drawingFigure 1
  • EP3232686B1 patent drawingFigure 2
  • EP3232686B1 patent drawingFigure 3

AI summary

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.