Echo Cancellation via Convolutive Blind Source Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing echo cancellation technologies fail to effectively remove acoustic echoes during doubletalk events, leading to impaired communication intelligibility and device functionality, especially in settings with simultaneous human or device speech.
Innovation Solution
The method employs convolutive blind source separation techniques, using a separating transfer function matrix to learn and adaptively adjust the acoustic echo transfer function, maximizing a criterion function to separate source signals and remove echoes even during doubletalk, by calculating a criterion function based on statistically independent output signals and employing gradient ascent optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional echo cancellation methods are used, then echo removal is achieved in single-talker scenarios, but communication intelligibility deteriorates during doubletalk events
Solution Approach 1:
The patent segments the mixed audio signal into multiple source signals using convolutive blind source separation. By decomposing the mixed signal x(n) into separate source signals s1(n) and s2(n) through independent component analysis, the system can identify and remove the echo path while preserving the desired speech signal, even during doubletalk events where both near-end and far-end speakers are active simultaneously.
Solution Approach 2:
The patent introduces an acoustic echo transfer function as an intermediary model that characterizes the echo path between the far-end speaker and near-end microphone. This transfer function serves as a mediator that enables the separation of echo components from the mixed signal without requiring direct access to the original source signals, allowing effective echo cancellation while maintaining speech intelligibility.
2Reliability
If adaptive echo cancellation is implemented, then echo removal performance improves, but computational complexity increases
Solution Approach 1:
The patent employs blind source separation techniques that enable the system to automatically identify and separate source signals without requiring prior knowledge or training data. The algorithm self-adapts to the acoustic environment by maximizing statistical independence of separated signals through information-theoretic criteria, eliminating the need for complex adaptive training phases while maintaining effective echo cancellation.
Solution Approach 2:
The patent changes the parameter space by using information-theoretic measures (mutual information, kurtosis) as optimization criteria instead of traditional least-squares approaches. This parameter transformation allows the system to achieve adaptive echo cancellation with reduced computational burden by exploiting statistical properties of the signals rather than requiring intensive iterative optimization of filter coefficients.
Data Source
AI summary
For canceling acoustic echoing, a processor receives audio signals comprising a speaker output and an ambient input. The processor further calculates separated output signals from mixed signals using a separating transfer function. The processor calculates a criterion function based on the separated output signals. In addition, the processor calculates an acoustic echo transfer function based on maximizing the a criterion function. The processor separates a source signal from the audio signal using the acoustic echo transfer function.