Blind Source Separation for Acoustic Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices face challenges in enhancing acoustic performance due to issues like double talk, reverb, and background noise, which degrade speech recognition and communication quality.
Innovation Solution
The integration of a blind source separation module with an acoustic echo canceller in devices to identify and remove acoustic echoes and background noise, using adaptive filtering and visual cues to determine the number of source signals and adjust filter adaptation during double-talk conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If acoustic echo cancellation is performed using traditional methods, then acoustic performance is maintained under normal conditions, but performance degrades significantly when double talk, reverb, or background noise is present
Solution Approach 1:
The patent segments the acoustic signal processing into distinct stages: blind source separation to isolate individual sound sources, followed by acoustic echo cancellation on the separated signals. This segmentation allows each processing stage to focus on specific aspects of the problem, improving overall reliability in noisy environments with double talk conditions
Solution Approach 2:
The patent introduces blind source separation as an intermediary processing stage between microphone input and acoustic echo cancellation. This intermediary separates mixed acoustic signals into distinct source components, enabling the echo canceller to operate more effectively even when multiple sources (including near-end speech) are present
2Adaptability or versatility
If the near end talker speaks into microphones while the loudspeaker plays far end signal, then voice communication functionality is enabled, but acoustic echo and double talk degrade speech recognition performance
Solution Approach 1:
The patent segments the audio signal into distinct source components using blind source separation, isolating the far end echo from the near end speech. This allows the system to maintain full-duplex voice communication while preserving speech recognition accuracy by processing separated signals through appropriate cancellation and recognition pipelines
Solution Approach 2:
The patent extracts and removes the acoustic echo component from the mixed microphone signals using blind source separation followed by echo cancellation. This extraction enables the near end talker's speech to be captured clearly despite the loudspeaker playing far end signal, maintaining both communication functionality and recognition precision
Data Source
AI summary
Techniques for utilizing blind source separation as a front-end to an acoustic echo canceller are described herein. The techniques include removing a first portion of an acoustic echo from an audio signal using blind source separation and a reference signal. The techniques then further remove a second portion of the acoustic echo using an acoustic echo canceller and the reference signal. Further, output of the blind source separation may be used to improve double-talk detection.