Voice Activity Detection via Subspace Correlation for Small Mic Arrays

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing voice enhancement technologies using the MVDR adaptive beamforming algorithm suffer from low precision in estimating noise covariance matrices, particularly when fewer than 5 microphones are used, leading to poor performance in devices with small microphone arrays like headsets.

Innovation Solution

A voice activity detection and enhancement system that calculates a voice presence probability by determining a signal subspace and a target subspace based on linear correlation between microphone signals, allowing for improved precision in identifying voice presence and enhancing audio signals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If the MVDR adaptive beamforming algorithm is used with a small quantity of microphones (less than 5), then the device complexity is reduced, but the measurement precision of noise covariance matrix estimation deteriorates sharply

Engineering Contradiction:

Improvequantity of microphonesVSAvoidnoise covariance matrix estimation precision

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent changes the parameter of voice presence probability estimation by introducing a subspace-based method that calculates the correlation between signal subspace and target voice subspace. This parameter transformation enables accurate noise covariance matrix estimation even with fewer than 5 microphones, resolving the contradiction between device complexity and measurement precision

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces the traditional mechanical/statistical approach of noise covariance estimation with a subspace-based mathematical method. By using signal subspace decomposition and correlation calculation, the system achieves high-precision estimation without relying on large microphone arrays, thus substituting a mathematical model for a physical constraint

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If the voice presence probability is estimated using existing VAD methods, then the system operation is simplified, but the measurement precision of voice presence probability deteriorates, leading to poor voice enhancement effect

Engineering Contradiction:

Improvevoice presence probability estimation processVSAvoidvoice presence probability accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent introduces a new dimension of analysis by transforming the voice presence detection problem from traditional time-domain or frequency-domain analysis to subspace-based correlation analysis. By calculating the linear correlation between signal subspace and target voice subspace, the system achieves high-precision voice presence probability estimation while maintaining operational simplicity

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS12475915B2Voice activity detection method and system, and voice enhancement method and system

Publication Date: 2025.11.18 SHENZHEN SHOKZ CO LTD

AI summary

A voice activity detection method and system and a voice enhancement method and system are provided. A voice presence probability of a target voice signal present in microphone signals may be determined by calculating a linear correlation between a signal subspace where the microphone signals are located and a target subspace where the target voice signal is located. The voice enhancement method and system may be used to calculate filter coefficients based on the voice presence probability, so as to perform voice enhancement on the microphone signals. The calculation accuracy of the voice presence probability is improved, and the voice enhancement effect is also improved.