Speech Noise Reduction via Adaptive Target Value Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional noise reducers distort speech signals and generate musical noise due to excessive noise reduction, particularly when dealing with nonstationary noise and background conversations, as they assume independence between noise and speech signals, leading to inefficient noise reduction and distortion.
Innovation Solution
A noise reducer system that estimates a target noise reduction value for each frequency band based on the speech signal, using a noise target value estimating part to adjust the noise reduction coefficient dynamically, ensuring that the noise is reduced without distorting the speech signal, even in environments with mixed stationary and nonstationary noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If the amplitude component of noise is subtracted from the input signal to reduce noise, then noise reduction is achieved, but the speech signal is distorted and musical noise is generated
Solution Approach 1:
The patent changes the parameter of noise reduction by introducing a target value that limits the maximum reduction level. Instead of simply subtracting noise amplitude, the system calculates a target value based on the ratio of noise to speech amplitude and uses this to control the noise reduction coefficient, preventing excessive reduction that causes distortion
Solution Approach 2:
The patent implements feedback by calculating the target value from the current signal characteristics (amplitude ratios of noise to speech) and using this target value to adjust the noise reduction coefficient. This feedback mechanism ensures that noise reduction adapts to the actual signal conditions and prevents over-reduction
2Object-affected harmful factors
If the noise reduction coefficient is increased to reduce more noise, then noise reduction effectiveness is improved, but speech distortion and musical noise increase
Solution Approach 1:
The patent introduces a target value parameter that dynamically controls the noise reduction coefficient. By calculating this target value based on the ratio of noise amplitude to speech amplitude in each frequency band, the system optimizes the reduction level to achieve effective noise removal while preventing the generation of musical noise and distortion
3Object-affected harmful factors
If noise reduction is applied to all frequency bands, then overall noise is reduced, but speech in frequency bands with speech present is distorted
Solution Approach 1:
The patent applies local quality by calculating and applying different target values for different frequency bands. Instead of uniform noise reduction across all frequencies, the system adapts the noise reduction coefficient to each frequency band's specific characteristics (noise-to-speech amplitude ratio), ensuring that speech-containing bands are preserved while noise-only bands are cleaned
Data Source
AI summary
Accepting the speech having the noise superimposed thereon and converting it into a signal on a time axis of the speech, an amplitude component of a speech for each predetermined frequency band of the converted signal on the frequency axis is calculated. Calculating a noise reduction coefficient, the noise component is reduced by multiplying the signal on the frequency axis of the original signal by the calculated noise reduction coefficient. By estimating the target value of the remaining noise for each frequency band, a signal on a frequency axis in which a signal corresponding to a frequency band of which target value estimated by the noise target value is larger than the value of the amplitude component of the signal on the frequency axis of which noise component is reduced is corrected to a signal corresponding to the target value is restored, into a signal on a time axis.