Multi-band Speech Separating Microphone Array Processor
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio communication systems struggle to effectively separate speech from ambient acoustic noise, particularly in environments where the speaker is moving, as they often rely on high-latency FFT algorithms and cannot combine source-attribute filtering with narrowband source tracking beam-forming for optimal signal-to-noise ratio reduction.
Innovation Solution
A digital signal processing system that uses a steering-controlled beam-former in combination with computational auditory scene analysis (CASA) to adaptively filter microphone signals, forming a spatial beam that tracks the speech source and suppresses residual noise, thereby improving speech separation and reducing latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If wideband multi-angle beam-forming is used to permit analysis of the entire ambient acoustic environment, then adaptability is improved, but signal-to-noise ratio reduction is worsened
Solution Approach 1:
The patent divides the acoustic signal processing into multiple frequency bands using filter banks, allowing different processing strategies for different frequency ranges. This segmentation enables the system to apply narrowband beam-forming where high selectivity is needed while maintaining overall environmental awareness through the multi-band structure.
Solution Approach 2:
The patent applies different processing characteristics to different frequency bands - narrowband beam-forming with high directivity for speech frequencies and wider bandwidth processing for other frequencies. This local differentiation optimizes signal-to-noise ratio in critical speech bands while preserving adaptability in other bands.
2Object-affected harmful factors
If narrowband source tracking beam-forming is used to provide high selectivity, then signal-to-noise ratio reduction is improved, but adaptability is worsened
Solution Approach 1:
The patent segments the audio spectrum into multiple frequency bands, applying narrowband beam-forming selectively to speech-relevant bands while using other processing approaches in non-speech bands. This maintains high selectivity where needed without sacrificing overall environmental analysis capability.
Solution Approach 2:
The multi-band beam-forming structure serves multiple functions simultaneously: it provides narrowband source tracking for speech enhancement while also maintaining awareness of the broader acoustic environment through the complete frequency band coverage, achieving both selectivity and adaptability.
3Measurement precision
If FFT algorithms are used for speech isolation, then speech separation is improved, but latency is worsened
Solution Approach 1:
The patent employs adaptive beam-forming that dynamically adjusts its parameters based on real-time acoustic conditions, allowing the system to optimize speech isolation performance adaptively rather than relying on fixed FFT-based processing, thereby reducing latency while maintaining precision.
Data Source
AI summary
A speech separating digital signal processing system and algorithms for implementing speech separation combine beam-forming with residual noise suppression, such as computational auditory scene analysis (CASA) using a beam-former that has a primary lobe steered toward the source of speech by a control value generated from an adaptive filter. An estimator estimates the ambient noise and provides an input to the residual noise suppressor, and a post-filter may be used to noise-reduce the output of the estimator using a time-varying filter that compares two or more outputs of the beam-former with a quasi-stationary model of the speech and ambient noise.