Speech Enhancement Based on Real-Time Noise Spectrum Estimation and Auditory Masking Effect
In order to improve the real-time feature of noise parameters estimation in the noisy speech signals, we propose one method of speech enhancement based on real-time noise spectrum estimation and auditory masking effect. This method makes use of the auditory masking threshold to modify the subtraction parameters. And we use the real-lime noise estimation to reduce the estimation error of the spectral subtraction noise on the basis of the modification. The experiment results show that the method presented in this paper has a better ability of tracking noise and also has a smaller noise estimation error than any of the other spectral subtractions, while effectively suppressing background noise, reducing musical noise and lowering speech distortion as well.
speech enhancement low SNR noise spectrum estimation auditory masking
Zhi Tao He-Ming Zhao Xiao-Jun Zhang Ji-Hua Gu
Dept. of Phys. Sci. and Tech. Dept. of Electron Soochow University SuZhou, China Dept. of Electron Soochow University SuZhou, China Dept. of Phys. Sci. and Tech. Soochow University SuZhou, China
国际会议
成都
英文
515-518
2010-12-17（万方平台首次上网日期，不代表论文的发表时间）