Dominant Speech Extraction Using Sub-Band Beamforming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current spatial noise cancellation techniques do not effectively exploit the spectral nature of audio signals and tend to favor either diffuse or directional noise, resulting in sub-optimal performance when combined with traditional single-channel noise cancellation.
Innovation Solution
A method that decomposes audio signals into low and high frequency sub-band signals, applies speech suppression beamforming to the low frequency signals to generate a noise estimate, uses noise cancellation on the primary audio signal, and combines the processed sub-band signals to enhance signal-to-noise ratio (SNR) for dominant speech extraction in the presence of both diffuse and directional noises.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional spatial noise cancellation techniques (filter-and-sum beamforming or super-directivity beamforming) are used, then the system can cancel noise, but the system does not effectively exploit the spectral nature of audio signals and tends to favor either diffuse or directional noise, resulting in sub-optimal performance
Solution Approach 1:
The patent divides the frequency spectrum into multiple sub-bands (low frequency and high frequency) and applies different beamforming techniques to each sub-band. Low frequency signals use speech suppression beamforming while high frequency signals use noise suppression beamforming. This segmentation allows the system to optimize performance for different noise types across different frequency ranges, resolving the contradiction between reliability and adaptability.
2Reliability
If cascaded traditional single channel noise cancellation is applied with spatial noise cancellation, then a complete noise cancellation solution is provided, but the spatial noise cancellation does not benefit from the traditional noise cancellation and vice-versa, resulting in sub-optimal performance
Solution Approach 1:
The patent merges spatial noise cancellation with traditional single-channel noise cancellation by integrating them at the sub-band level. The system applies both speech suppression beamforming and noise cancellation algorithms within the same processing framework, allowing the two techniques to work synergistically rather than independently. This combination improves overall performance efficiency while providing a complete noise cancellation solution.
3Measurement precision
If beamforming techniques are applied to extract dominant speech, then the system can separate speech from noise, but the system does not effectively exploit the spectral nature of audio signals
Solution Approach 1:
The patent applies different processing qualities to different parts of the frequency spectrum. Low frequency sub-bands receive speech suppression beamforming which exploits the spectral characteristics of speech in that range, while high frequency sub-bands receive noise suppression beamforming. This local optimization of processing quality for different frequency regions improves speech extraction accuracy while managing processing complexity through targeted approaches.
Data Source
AI summary
A method of dominant speech extraction is provided that includes acquiring a primary audio signal from a microphone and at least one additional audio signal from at least one additional microphone, wherein the acquired audio signals include speech and noise, decomposing each acquired audio signal into a low frequency sub-band signal and a high frequency sub-band signal, applying speech suppression beamforming to the low frequency sub-band signals to generate a reference channel having an estimate of noise in the low frequency sub-band signals, applying noise cancellation to the low frequency sub-band signal of the primary audio signal using the reference channel to generate a first signal having a low frequency estimate of the speech, applying noise suppression beamforming to the high frequency sub-band signals to generate a second signal having a high frequency estimate of the speech, and combining the first and second signals to generate a full-band audio signal.