Adaptive Speech Enhancement via Dynamic Spectral Subtraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional spectral subtraction methods in speech enhancement often generate 'musical noise', affecting the intelligibility and naturalness of speech signals, especially in environments with varying noise levels and signal-to-noise ratios.
Innovation Solution
An adaptive speech enhancement method that adjusts spectral subtraction parameters based on the power spectrum features of user speech and environmental noise, using predicted power spectra to optimize the subtraction process, thereby improving noise reduction and signal quality across a wide range of signal-to-noise ratios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional spectral subtraction is performed to eliminate noise in the sound signal, then noise reduction is achieved, but musical noise is generated in the denoised speech signal, affecting intelligibility and naturalness
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting spectral subtraction parameters (over-subtraction factor α and spectrum order β) based on the estimated signal-to-noise ratio and power spectrum characteristics. Instead of using fixed parameters, the system adapts them in real-time to match the acoustic environment, which prevents musical noise generation while maintaining effective noise reduction across varying signal-to-noise ratio conditions
Solution Approach 2:
The patent implements dynamics by making the spectral subtraction process adaptive rather than static. The system continuously estimates the power spectrum of noise and speech signals, calculates the signal-to-noise ratio, and dynamically adjusts the subtraction parameters accordingly. This dynamic adaptation allows the noise reduction system to respond to changing acoustic conditions without generating musical noise artifacts
2Adaptability or versatility
If spectral subtraction parameters are fixed, then the processing is simple, but the method is not applicable to a wide signal-to-noise ratio range and produces poor noise reduction performance
Solution Approach 1:
The patent employs feedback by using the estimated power spectrum of the noise signal and the speech signal containing noise to continuously monitor the signal-to-noise ratio. Based on this feedback information, the system automatically adjusts the spectral subtraction parameters (α and β) to optimize noise reduction performance across different signal-to-noise ratio ranges, achieving wide adaptability without manual intervention
Solution Approach 2:
The system performs self-service by autonomously estimating the noise power spectrum, calculating the signal-to-noise ratio, and adjusting the spectral subtraction parameters without requiring external control or manual tuning. The adaptive algorithm automatically adapts to the acoustic environment and optimizes the denoising process, simplifying the overall system operation while maintaining high performance across varying conditions
Data Source
AI summary
A speech enhancement method includes determining a first spectral subtraction parameter based on a power spectrum of a speech signal containing noise and a power spectrum of a noise signal, determining a second spectral subtraction parameter based on the first spectral subtraction parameter and a reference power spectrum, and performing, based on the power spectrum of the noise signal and the second spectral subtraction parameter, spectral subtraction on the speech signal containing noise, where the reference power spectrum includes a predicted user speech power spectrum and/or predicted environmental noise power. Regularity of a power spectrum feature of a user speech of a terminal device and/or regularity of a power spectrum feature of noise in an environment in which a user is located are considered.