Yasser Ghanbari - Department of Electrical Engineering, University of Mazandaran, Babol, Iran
Mohammad Reza Karami - Department of Electrical Engineering, University of Mazandaran, Babol, Iran
Mojtaba Lotfizad - Department of Electrical Engineering, Tarbiat Modares University, Tehran, Iran
Hosein Sameti - Department of Computer Engineering, Sharif University of Technology, Tehran, Iran

In this paper, we propose a new approach for speech enhancement. The proposed method for removing the noise components is a combination of two methods: Wavelet de-noising and spectral subtraction. The idea is to apply the spectral subtraction to the wavelet approximations and details coefficients. The spectral subtraction algorithm is modified through the new algorithm by modification of some parameters including a new parameter for spectral subtraction in unvoiced speech frames which is introduced and the existing power factor in spectral subtraction method which is improved. Also, for reduction of musical noise, we propose to use iterative filtering of noises which can remove the musical noise, sufficiently. Experimental results on a large number of noisy signals and comparisons between four algorithms demonstrate that the proposed speech enhancement algorithm is promising.

Speech enhancement, spectral subtraction, wavelet denoising