Array model interpolation and subband iterative adaptive filters applied to beamforming-based acoustic echo...

Area of Science:

• Signal Processing
• Acoustics
• Speech Technology

Background:

• Acoustic echo cancellation (AEC) is crucial for teleconferencing and communication systems.
• Traditional AEC methods struggle with real-world acoustic environments and dynamic conditions like double-talk.
• Limitations exist in accurately modeling array steering vectors and achieving rapid convergence in adaptive filters.

Purpose of the Study:

• To develop and evaluate enhanced strategies for acoustic echo cancellers (AECs).
• To improve echo cancellation performance and speech quality in challenging acoustic scenarios.
• To introduce novel techniques for array modeling and adaptive filter convergence.

Main Methods:

• Utilized a fixed beamformer (FBF) for near-end speaker focus and far-end echo suppression.
• Developed an experimental procedure to interpolate practical array models from measured frequency responses.
• Implemented subband (SB) filtering with polyphase implementation and combined a generalized sidelobe canceller (GSC) with AEC, incorporating an internal iteration (IIT) procedure for efficient adaptive filter convergence.

Main Results:

• The proposed GSC-SB-AEC-IIT approach demonstrated the highest echo return loss enhancement (ERLE).
• Objective and subjective tests confirmed no degradation in speech quality (PESQ, listening tests).
• The method proved effective even in double-talk scenarios, maintaining high ERLE.

Conclusions:

• The GSC-SB-AEC-IIT approach represents a significant advancement in acoustic echo cancellation.
• This method offers robust performance, maintaining high echo suppression and speech intelligibility.
• The developed techniques are suitable for real-world applications requiring high-quality audio transmission.