Multi-microphone Signal Enhancement via Adaptive Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-microphone audio processing techniques often lose or distort spatial information about sound sources, leading to noise-reduced audio signals that lack coherence and accuracy in phase and magnitude relationships between microphone signals.
Innovation Solution
The use of an adaptive filter to generate predicted microphone signals based on correlated signal portions from multiple microphones, preserving spatial information and reducing noise content while maintaining high coherence, by selecting a reference microphone signal and convolving it with adaptive filter parameters determined through optimization algorithms like LMS.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If multiple microphones are used to generate noise-reduced audio signals, then noise content is reduced, but spatial information about sound sources is lost or distorted
Solution Approach 1:
The system uses adaptive filtering where the filter parameters are continuously adjusted based on feedback from comparing the reference microphone signal with signals from other microphones. This feedback mechanism allows the system to learn and adapt to the spatial characteristics of sound sources while reducing noise, thereby preserving spatial information rather than losing it.
Solution Approach 2:
The patent changes the parameters of the adaptive filter (such as filter coefficients) dynamically based on the statistical properties of the microphone signals. By adapting these parameters in real-time, the system can maintain accurate spatial representation while achieving noise reduction through optimal filtering.
2Object-affected harmful factors
If adaptive filtering is applied to reduce noise, then noise content is reduced, but coherence and accuracy in phase and magnitude relationships deteriorate
Solution Approach 1:
The adaptive filter uses feedback from the reference microphone signal to continuously adjust its parameters. This ensures that the filtering process maintains coherence and accuracy in phase and magnitude relationships by constantly adapting to the actual signal characteristics rather than applying fixed transformations that would distort these relationships.
Solution Approach 2:
The system transitions from static filtering to dynamic adaptive filtering where the filter characteristics change over time based on signal conditions. This dynamic adaptation preserves the temporal and spectral coherence of the signals while achieving noise reduction, as the filter responds to actual signal variations rather than applying predetermined transformations.
3Object-affected harmful factors
If noise reduction processing is applied to microphone signals, then noise content is reduced, but spatial cues and directional components are lost or distorted
Solution Approach 1:
The system employs feedback from multiple microphone signals including a reference signal to adaptively adjust filtering parameters. This feedback mechanism ensures that spatial cues and directional components are preserved during noise reduction, as the adaptive filter learns the spatial characteristics of both desired signals and noise from the multi-microphone input.
Solution Approach 2:
The patent segments the noise reduction process into multiple adaptive filters, each handling different spatial components or frequency ranges. By dividing the processing into separate adaptive channels that can be independently optimized, the system preserves spatial cues while achieving effective noise reduction in each segment.
Data Source
AI summary
Microphone signals are received from microphones of a computer device. Each microphone signal of the microphone signals is acquired by a respective microphone of the microphones. A previously unselected microphone is selected from the microphones as a reference microphone, which generates a reference microphone signal. An adaptive filter is used to create, based on microphone signals of the microphones other than the reference microphone, predicted microphone signals for the reference microphone. Based on the predicted microphone signals for the reference microphone, an enhanced microphone signal is outputted for the reference microphone. The enhanced microphone signal may be used as microphone signal for the reference microphone in subsequent audio processing operations.