Audio Signal Breath Sound Suppression via Frequency Band Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In audio communication scenarios, unwanted breath sounds are often captured by microphones and can be loud and disturbing, degrading user experience, as existing systems either fail to suppress them effectively or enhance their level unintentionally.
Innovation Solution
A method and system that detect the type of audio signal, applying a suppression gain to mitigate loud breath sounds while maintaining an appropriate level for voice sounds, using features like spectral difference, signal-to-noise ratio, and spectral variance to classify and process the audio signal frame by frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If automatic gain control is applied to audio signals, then the overall sound level is adjusted, but breath sounds are unintentionally enhanced along with voice sounds
Solution Approach 1:
The audio signal is divided into multiple frequency bands, and each band is processed independently with different gain values. Voice frequencies receive gain adjustment while breath frequencies are excluded, allowing selective amplification that enhances speech without unintentionally enhancing breath sounds.
Solution Approach 2:
Different gain values are applied to different frequency regions based on their characteristics. The system identifies frequency ranges corresponding to voice and assigns them higher gain, while frequency ranges corresponding to breath sounds are assigned lower or zero gain, creating local quality differentiation in the audio processing.
2Object-affected harmful factors
If breath sounds are suppressed using traditional noise suppression methods, then some breath noise is reduced, but voice quality is also degraded
Solution Approach 1:
The frequency spectrum is segmented into multiple bands, and the system selectively applies suppression only to bands identified as containing breath sounds. Voice bands are excluded from suppression, thereby reducing breath noise while preserving voice quality without degrading speech intelligibility.
Solution Approach 2:
The suppression operation is applied locally only to frequency regions where breath sounds are detected, rather than uniformly across the entire audio spectrum. This localized approach ensures that breath suppression occurs only where needed while leaving voice frequencies untouched, maintaining high voice quality.
3Measurement precision
If the microphone is placed close to the user's mouth, then voice capture is improved, but breath sounds become louder and more disturbing
Solution Approach 1:
The audio signal captured by the close-placed microphone is divided into frequency bands, with separate processing for voice and breath frequencies. This allows the system to maintain high voice capture quality while selectively suppressing the louder breath sounds that result from the close microphone placement.
Solution Approach 2:
The system applies different processing characteristics to different frequency regions of the captured audio signal. Frequency regions corresponding to voice are preserved with high fidelity, while frequency regions corresponding to breath sounds are attenuated, creating local quality differentiation that compensates for the close microphone placement.
Data Source
Figure 1
Figure 2~4
Figure 5~7A
AI summary
Example embodiments disclosed herein relate to audio signal processing. A method of processing an audio signal is disclosed. The method includes detecting, based on a power distribution of the audio signal, a type of content of a frame of the audio signal, generating a first gain based on a sound level of the frame for adjusting the sound level, processing the audio signal by applying the first gain to the frame; and in response to the type of content being detected to be a breath sound, generating a second gain for mitigating the breath sound and processing the audio signal by applying the second gain to the frame. Corresponding system and computer program product are also disclosed.