Microphone Array Source Separation for Consistent Voice Recording
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound source recording devices struggle to maintain a consistent recording level for target sound sources in environments with varying ambient noise and sound source characteristics, leading to suboptimal sound recognition, especially in indoor and outdoor settings where multiple noise sources are present.
Innovation Solution
The implementation of a sound source recording method using independent component analysis (ICA) and geometric source separation (GSS) technologies to separate target sound sources from received signals, with a gain estimation formula (L=Lref*dref^2/d^2*G=B*L) and adaptive gain control based on sound source distance, allowing for real-time adjustment of microphone gain to maintain a predetermined recording level.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gain control is performed based only on sound source signal intensity, then the sound source signal level is adjusted, but the recorded voice level varies due to ambient noise changes
Solution Approach 1:
The patent segments the sound source signal from ambient noise by using multiple microphones to capture spatial information. The sound source signal is separated into a target component and a noise component through spatial filtering, allowing independent processing of each component to achieve stable voice recording despite ambient noise variations.
Solution Approach 2:
The patent introduces spatial filtering as an intermediary process between the raw sound source signal and the final recorded output. By using microphone array processing and spatial filtering techniques, the system mediates between the varying sound source intensity and the desired consistent recording level, eliminating the direct coupling that causes instability.
2Adaptability or versatility
If gain is adjusted based on sound source distance, then distant sounds are amplified, but close sounds may be overamplified causing clipping
Solution Approach 1:
The patent implements dynamic gain control that adapts to both distance and instantaneous sound intensity. The gain factor is calculated as a function of estimated distance and current sound level, allowing the system to dynamically adjust between amplification for distant sources and attenuation or clamping for close sources, preventing overamplification while maintaining adaptability.
Solution Approach 2:
The patent changes the gain parameter based on multiple variables including estimated sound source distance and current sound intensity level. By making the gain a function of multiple parameters rather than a fixed value or single-parameter control, the system achieves uniform sound level output across varying distances and intensities.
3Measurement precision
If microphone gain is increased to capture distant sound sources, then distant voices are captured, but ambient noise is also amplified
Solution Approach 1:
The patent extracts the target sound source from the mixed signal by using spatial information from multiple microphones. Through spatial filtering and source separation techniques, the system takes out the desired voice signal from the combination of voice and ambient noise, allowing selective amplification of the extracted source without amplifying the remaining noise component.
Solution Approach 2:
The patent applies different processing qualities to different components of the sound field. The target sound source receives amplification and enhancement processing, while the ambient noise component receives different or no processing. This local differentiation of processing quality allows distant sources to be captured while ambient noise remains suppressed.
Data Source
AI summary
Disclosed herein is a sound source recording apparatus and method adaptable to an operating environment, which can record a target sound source at a predetermined level without being affected by characteristics of the sound source or ambient noise. A target sound source is separated from a sound source signal received through an array of microphones and a recording sound pressure level and a gain are estimated using a reference sound pressure level and a reference distance for the target sound source, thereby controlling or adjusting the gain of the microphones.