Dynamic Low Frequency Noise Suppression for Speech Clarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional communication systems using Wiener filters struggle to effectively suppress non-stationary noise bursts due to their dependency on Signal-to-Noise Ratio and unreliable detection of noise power spectral density, leading to inadequate suppression of low frequency noise events like road bumps, which can interfere with speech clarity.
Innovation Solution
A dynamic low frequency noise suppression method that uses adjustable windows to compute a dampening level based on signal peak information, ensuring that non-speech audio events such as road bumps are suppressed without affecting speech components, by forming and adjusting first and second windows to maximize dampening while preserving speech formants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Wiener filters are used to suppress noise, then stationary noise suppression is improved, but non-stationary noise bursts cannot be suppressed adequately due to slow response
Solution Approach 1:
The patent implements dynamic window size adjustment where the analysis window length is adapted based on the detected noise burst duration. For short non-stationary noise bursts, shorter windows are used to achieve quick response, while for stationary noise, longer windows maintain suppression reliability. This dynamic adaptation resolves the contradiction between fast response for bursts and reliable suppression for stationary noise.
Solution Approach 2:
The system changes the window size parameter dynamically based on noise characteristics. When a noise burst is detected, the window size is reduced to improve temporal resolution and response speed. For stationary noise conditions, larger window sizes are used to maintain frequency resolution and suppression reliability, thus resolving the speed-reliability tradeoff.
2Measurement precision
If conventional noise suppression is used, then noise power spectral density estimation is performed, but detection of relatively short bursts is unreliable
Solution Approach 1:
The patent uses dynamic window sizing where shorter windows are employed specifically for detecting short noise bursts. This allows the system to maintain measurement precision for PSD estimation while improving reliability for short burst detection, as the shorter windows provide better temporal resolution to capture brief events.
Solution Approach 2:
The patent segments the noise analysis into different temporal regions using variable window sizes. Short windows are applied to segments containing brief noise bursts, while longer windows are used for stationary noise segments. This segmentation allows reliable detection of short bursts without compromising overall PSD estimation accuracy.
3Object-affected harmful factors
If low frequency noise suppression is applied, then noise events like road bumps are suppressed, but speech formants may be suppressed along with noise
Solution Approach 1:
The patent applies different suppression strengths to different frequency regions within the low frequency range. By analyzing the spectral content and identifying speech formants, the system applies targeted suppression only to noise components while preserving speech formant regions. This local differentiation resolves the contradiction between noise suppression and speech preservation.
Solution Approach 2:
The system uses feedback from spectral analysis to dynamically adjust suppression levels. By continuously monitoring the input signal for speech formant characteristics, the feedback mechanism prevents suppression of speech components while maintaining effective noise suppression in non-speech regions, thus resolving the harmful effects contradiction.
Data Source
AI summary
Methods and apparatus for dynamically suppressing low frequency non-speech audio events, such as road bumps, without suppressing speech formants. In exemplary embodiments of the invention, maximum powers in first and second windows are computed and used to determine whether dampening should be applied, and if so, to what extent.