Acoustic Echo Cancellation via Frequency-Domain Blind Source Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic echo cancellation techniques, such as the normalized least-mean squares (NLMS) method, are less accurate during double-talk conditions in full duplex communication, leading to reduced accuracy and slower convergence rates, especially when users concurrently speak into microphones.
Innovation Solution
The method iteratively adapts the estimate of the acoustic echo using blind source separation (BSS) operations to increase statistical independence between signals in the frequency domain, regardless of the presence of voices in the sound waves, thereby improving echo cancellation and convergence rates without requiring a double-talk detector or suspending adjustments during double-talk conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If NLMS technique is used for acoustic echo cancellation, then the system can operate in full duplex mode, but the accuracy deteriorates during double-talk conditions
Solution Approach 1:
The patent segments the echo cancellation problem into frequency-domain processing, dividing the signal into multiple frequency bins and applying independent processing to each bin. This allows the system to handle double-talk conditions more effectively by treating different frequency components separately, improving accuracy during concurrent speech scenarios.
Solution Approach 2:
The patent transforms the time-domain NLMS algorithm into a frequency-domain implementation, changing the domain parameter from time to frequency. This transformation enables better handling of non-stationary signals during double-talk by allowing frequency-selective processing and adaptive filtering in the frequency domain, thereby maintaining accuracy while preserving full duplex capability.
2Adaptability or versatility
If NLMS technique is used for acoustic echo cancellation, then the system can adapt to changing acoustic environments, but the convergence rate slows down during double-talk conditions
Solution Approach 1:
The patent implements dynamic adaptation by using frequency-domain processing that can independently adjust to different acoustic conditions at each frequency bin. The system dynamically adapts to changing environments while maintaining fast convergence through frequency-selective filtering, allowing rapid response to environmental changes without the slow convergence problem of traditional time-domain methods during double-talk.
Solution Approach 2:
The patent transitions from time-domain to frequency-domain processing, adding a frequency dimension to the adaptation process. This dimensional change enables parallel adaptation across multiple frequency bins, significantly accelerating convergence rates while maintaining adaptability to changing acoustic environments, especially during double-talk conditions where time-domain methods struggle.
3Device complexity
If traditional echo cancellation methods are used, then the system structure remains simple, but misalignment occurs during double-talk conditions
Solution Approach 1:
The patent introduces frequency-domain processing as an intermediary layer between the input signal and the echo cancellation output. This intermediary transformation enables precise alignment by allowing independent phase and magnitude adjustment at each frequency bin, correcting misalignment issues that occur during double-talk without significantly increasing overall system complexity.
Data Source
AI summary
In response to a first signal, a first sound wave is output. A second sound wave is received that includes an acoustic echo of the first sound wave. In response to the second sound wave, a second signal is output that cancels an estimate of the acoustic echo. The estimate of the acoustic echo is iteratively adapted to increase a statistical independence between the first and second signals, irrespective of whether a first voice is present in the first sound wave, and irrespective of whether a second voice is present in the second sound wave.