Beamforming Adaptation During Speech Attacks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing beamforming audio capture systems face challenges in reverberant environments and distances beyond the reverberation radius, leading to suboptimal performance due to difficulties in distinguishing between echoes and diffuse background noise, slow adaptation, and increased noise sensitivity.
Innovation Solution
An audio capture apparatus with a beamformer, adapter, and controller that detects speech attacks to adapt beamform parameters within predetermined time intervals, focusing on direct and early reflections while disregarding late reflections, thereby improving adaptation and noise resilience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If continuous adaptation of beamform parameters is performed, then the system can track audio sources, but the adaptation is slow and ineffective in reverberant environments with late reflections
Solution Approach 1:
The patent applies periodic action by restricting beamformer adaptation to specific time intervals called adaptation time intervals, which occur periodically during speech attacks. Instead of continuous adaptation, the system performs adaptation only during these designated intervals when the direct sound path is dominant, thereby reducing the effective adaptation time while maintaining tracking capability.
Solution Approach 2:
The system performs preliminary action by detecting speech attacks in advance and using this detection to trigger adaptation time intervals. The speech attack detection serves as a preliminary condition that prepares the system to adapt beamform parameters at the optimal moment when direct sound is dominant, before late reflections can interfere with the adaptation process.
2Quantity of substance
If beamforming adapts to all received signals including late reflections, then the system captures more audio energy, but speech capture accuracy deteriorates due to inability to distinguish echoes from desired speech
Solution Approach 1:
The patent applies segmentation by dividing the received audio signal into different temporal components: direct sound, early reflections, and late reflections. The beamformer adaptation is selectively applied only to the direct sound and early reflection components during speech attack intervals, while excluding late reflections. This segmentation allows the system to capture sufficient audio energy from direct paths while maintaining accuracy by excluding problematic echoed components.
3Reliability
If the system processes all audio signals continuously, then no speech is missed, but noise sensitivity increases in reverberant environments
Solution Approach 1:
The system applies periodic action by processing audio signals selectively during speech attack intervals rather than continuously. The speech attack detection triggers periodic processing windows where beamforming adaptation and noise suppression are applied. This periodic processing maintains speech detection reliability by ensuring speech is captured during attacks while reducing noise sensitivity by avoiding processing during intervals when only reverberant noise is present.
Data Source
AI summary
An audio capture apparatus comprises a first beamformer (303) which is arranged to generate a beamformed audio output signal. An adapter (305) adapts beamform parameters of the first beamformer and a detector (307) detects an attack of speech in the beamformed audio output signal. A controller (309) controls the adaptation of the beamform parameters to occur in a predetermined adaptation time interval determined in response to the detection of the attack of speech. The beamformer (303) may generate noise reference signal(s) and the detector (309) may be arranged to detect the attack of speech in response to a comparison of a signal level of the beamformed audio output signal relative to a signal level of the at least one noise reference signal.