Wideband Doubletalk Detection for Acoustic Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Acoustic echo cancellation systems struggle with suboptimal performance during doubletalk conditions, particularly when microphones and loudspeakers are close, leading to unwanted attenuation of near-end speech and introduction of mechanical vibrations, which degrade audio quality.
Innovation Solution
Systems and methods that override certain gains of echo-cancelled audio signal subbands, compress remote audio signals, and apply makeup gain based on detected doubletalk conditions to optimize audio quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a non-linear processor fully suppresses residual echo during doubletalk conditions, then echo is reduced, but near-end speech is unintentionally attenuated
Solution Approach 1:
The patent applies local quality by treating different frequency subbands differently during doubletalk conditions. Instead of uniformly suppressing all frequencies, the system identifies and preserves subbands containing near-end speech while suppressing subbands containing echo, thereby locally optimizing quality for each subband based on its content characteristics
Solution Approach 2:
The system dynamically changes the suppression parameter (gain) of the non-linear processor based on doubletalk detection. When doubletalk is detected, the system adjusts the suppression level selectively per subband rather than applying a fixed high suppression level, allowing adaptation between echo suppression and speech preservation modes
2Volume of moving object
If microphones and loudspeaker are placed close together, then device size is reduced, but mechanical vibrations increase and degrade audio quality
Solution Approach 1:
The patent extracts and removes the harmful mechanical vibration components from the audio signal through signal processing. The system detects vibrations caused by close proximity placement and selectively removes these artifacts from the transmitted audio, allowing close placement without sacrificing audio quality
Solution Approach 2:
The system introduces an intermediary processing stage (vibration detection and removal mechanism) between the physical close placement of components and the final audio output. This intermediary layer detects and compensates for the harmful effects of close placement, enabling compact design while maintaining audio quality
3Illumination intensity
If far end audio energy is increased for clear transmission, then far-end speech is improved, but near-end speech detection becomes more difficult
Solution Approach 1:
The system dynamically changes the energy level parameter of far-end audio based on detection of near-end speech activity. When near-end speech is detected, the system reduces far-end audio energy to prevent masking effects, and increases it when no near-end speech is present, thereby optimizing both far-end clarity and near-end detection accuracy
Solution Approach 2:
The patent implements feedback by continuously monitoring the microphone signal for near-end speech and using this information to adjust the far-end audio output level. This closed-loop control ensures that far-end audio energy is optimized while avoiding interference with near-end speech detection and transmission
Data Source
AI summary
Acoustic echo cancellation systems and methods are provided that improve the quality of the audio transmitted from by an audio device a near end to a far end when a doubletalk condition is present, including allowing certain subbands of an echo-cancelled signal to be less attenuated by overriding certain gains of subbands of the echo-cancelled audio signal in a non-linear processor, and compressing and applying makeup gain to a remote audio signal.