Multi-Channel Noise Suppression Gain Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise suppression methods in mobile telephone conversations are ineffective in responding to fast changes in signals due to infrequent updates of noise suppression parameters, leading to less effective noise reduction, and frequent updates result in annoying musical noise artifacts.
Innovation Solution
A method that estimates speech and noise levels within specific frequency bands, computes noise suppression gains based on these levels, and applies these gains to generate an output channel, using a combination of adaptive filtering, blind source separation, and post-processing techniques to suppress noise while minimizing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If noise suppression parameters are updated infrequently, then computational complexity is reduced, but noise suppression effectiveness deteriorates due to inability to respond to fast changes in signals
Solution Approach 1:
The patent segments the noise suppression process into multiple frequency bands (e.g., 8 frequency bands from 100Hz to 4kHz). By processing each band independently with its own gain factor, the system can respond to fast changes in specific frequency regions without requiring full recalculation of all parameters, thus reducing computational complexity while maintaining noise suppression effectiveness.
Solution Approach 2:
The patent implements dynamic update of noise suppression parameters by continuously tracking speech and noise levels in each frequency band. The system adaptively adjusts gain factors based on real-time signal characteristics, allowing it to respond to fast changes in signals. The update mechanism balances responsiveness with computational efficiency by updating parameters only when necessary.
2Reliability
If noise suppression parameters are updated frequently, then noise suppression effectiveness is improved, but musical noise artifacts increase
Solution Approach 1:
The patent applies preliminary smoothing to the estimated speech and noise levels before computing gain factors. By using smoothed estimates rather than raw instantaneous values, the system avoids rapid fluctuations in gain that cause musical noise artifacts. This preliminary processing step prepares the data in a way that prevents harmful artifacts while maintaining noise suppression effectiveness.
Solution Approach 2:
The patent implements a feedback mechanism where the output of the noise suppression process is monitored and used to adjust subsequent processing. The system detects musical noise artifacts and reduces the update frequency or magnitude of gain factors when such artifacts are detected, creating a closed-loop control that prevents harmful artifacts while maintaining effective noise suppression.
3Device complexity
If single-channel noise suppression is used, then device complexity is reduced, but noise suppression performance is insufficient in multi-talker scenarios
Solution Approach 1:
The patent transitions from single-channel to multi-channel processing by utilizing multiple microphones (e.g., far-end and near-end microphones). This dimensional expansion allows the system to separate speech signals from different talkers spatially and apply channel-specific noise suppression. The multi-channel approach maintains manageable complexity while significantly improving performance in multi-talker scenarios.
Data Source
AI summary
In response to a first envelope within a kth frequency band of a first channel, a speech level within the kth frequency band of the first channel is estimated. In response to a second envelope within the kth frequency band of a second channel, a noise level within the kth frequency band of the second channel is estimated. A noise suppression gain for a time frame n is computed in response to the estimated speech level for a preceding time frame, the estimated noise level for the preceding time frame, the estimated speech level for the time frame n, and the estimated noise level for the time frame n. An output channel is generated in response to multiplying the noise suppression gain for the time frame n and the first channel.