Correlation-Maximizing Wind Noise Removal With Sound Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wind noise suppression technologies struggle with accurate detection and removal, leading to distorted sound signals and degraded voice call quality, especially when sound and wind noise occur simultaneously.
Innovation Solution
A device and method that utilize multiple microphones, correlation maximizing filters, sound source separators, and residual wind noise removers to separate sound signals and wind noise without a separate detector, using weight vectors to maximize and minimize correlations between channels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a separate wind noise detector is used to detect wind noise, then wind noise can be identified, but the accuracy of detection deteriorates when sound signal and wind noise occur simultaneously, leading to signal distortion
Solution Approach 1:
The patent separates the mixed sound signal into distinct components (wind noise and sound signal) using sound source separation techniques. Instead of detecting wind noise separately in a mixed signal, the system divides the signal space into multiple sources, allowing accurate identification of wind noise without distorting the sound signal.
Solution Approach 2:
The sound source separation module performs multiple functions simultaneously: it separates wind noise from sound signals, enhances sound signals, and suppresses background noise all through a unified separation process, eliminating the need for separate detection and processing stages.
2Object-affected harmful factors
If wind noise is removed aggressively, then wind noise removal performance improves, but the sound signal becomes distorted
Solution Approach 1:
The patent segments the sound signal into multiple source components through sound source separation, identifying wind noise as a distinct source. This allows targeted removal of wind noise while preserving the integrity of sound signal components, avoiding the distortion that occurs with aggressive single-stage filtering.
Solution Approach 2:
The sound source separation module acts as an intermediary between the mixed signal and the final output. It first separates and identifies wind noise components, then allows selective processing and removal, serving as a mediator that protects the sound signal from direct distortion while enabling effective wind noise suppression.
3Device complexity
If sound source separation is performed without a separate wind noise detector, then the system complexity is reduced, but the ability to accurately identify and remove wind noise deteriorates
Solution Approach 1:
The sound source separation module performs multiple functions that would traditionally require separate components: it separates wind noise from sound signals, identifies wind noise sources, and enables subsequent removal processing. This multi-functional approach reduces system complexity while maintaining or improving wind noise identification accuracy through advanced separation algorithms.
Solution Approach 2:
The patent merges the wind noise detection function into the sound source separation module. Instead of having a separate detector, the separation process itself identifies and characterizes wind noise sources, combining detection and separation functions into a unified operation that reduces overall system complexity.
Data Source
AI summary
A device for removing wind noise is provided. The device includes: processors configured to execute more instructions to control: a correlation maximizing filter to output a third sound signal to a first channel based on a first sound signal, a second sound signal and a first weight vector, and output a fourth sound signal to a second channel based on the first sound signal, the second sound signal and a second weight vector; a sound source separator to receive the third and fourth sound signals through the first and second channels, and perform a sound source separation to output a fifth and sixth sound signals to the first and second channels; a residual wind noise remover to receive the fifth and sixth sound signals, and perform a residual wind noise removal based on the fifth and sixth sound signals to generate an output signal.