Voice Activity Detector for Acoustic Echo Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional acoustic echo cancellation systems, especially in mobile phone speakerphones, face significant degradation due to severe nonlinear distortion, leading to half-duplex communication and reduced quality, as they fail to distinguish between far-end echo and near-end speech effectively.
Innovation Solution
Implementing near-end voice activity detection based on both near-end input power and expected acoustic echo cancellation gain to enhance echo suppression, allowing for partial-duplex communication even under severe nonlinear distortion conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional echo suppression is applied to reduce AEC residual echo, then far-end echo is attenuated, but near-end speech is also attenuated causing half-duplex communication
Solution Approach 1:
The patent changes the parameter used for voice activity detection from simple near-end signal power to a combined parameter that includes both near-end signal power and expected AEC gain. This allows the system to distinguish between near-end speech and AEC residual echo by comparing the actual near-end power with the expected power after echo cancellation, enabling full-duplex communication while suppressing far-end echo
Solution Approach 2:
The patent introduces an intermediary calculation (expected AEC gain) that mediates between the near-end speech signal and the AEC residual echo. By calculating what the near-end power should be after echo cancellation and comparing it with actual measurements, the system can identify when near-end speech is present even in the presence of significant AEC residual echo, preventing unnecessary attenuation
2Object-affected harmful factors
If AEC is used to cancel acoustic echo, then linear echo is reduced, but severe nonlinear distortion remains causing detection failure
Solution Approach 1:
The patent applies partial action by using a two-stage approach: first AEC handles the linear echo cancellation, then the improved VAD with expected AEC gain handling deals with the remaining nonlinear distortion. This partial application of different techniques allows the system to address both linear and nonlinear components separately, maintaining detection accuracy despite severe nonlinear distortion
3Object-affected harmful factors
If echo suppression attenuates both far-end echo and near-end speech, then far-end echo becomes inaudible, but communication quality significantly degrades
Solution Approach 1:
The patent implements feedback by continuously monitoring the near-end signal power and comparing it with the expected power after echo cancellation. This feedback mechanism allows the echo suppressor to adaptively adjust attenuation levels based on whether near-end speech is detected, ensuring far-end echo suppression while maintaining near-end speech quality and enabling full-duplex operation
Data Source
AI summary
Acoustic echo control for hands-free phone has acoustic echo cancellation and echo suppression with a voice activity detection for the echo suppression based on near-end input power together with an estimate for acoustic echo cancellation gain.