Denoising speech by notch filter and wavelet thresholding in real time
Mehmet Alper Oktar, Mokhtar Nibouche, Yusuf Baltacı · 2016
Real time speech signal noise reduction is vital to improve the speech quality and intelligibility for human listener. Wavelet thresholding gives good results in denoising speech corrupted with white Gaussian noise. However, if the speech is corrupted with both fixed frequency and white Gaussian noise, wavelet thresholding fails to generate high speech quality due to the fact that the fixed frequency noise components corrupt the choice of the threshold. This paper presents a new speech enhancement approach utilising a combination of wavelet thresholding and notch filter. Experiment results show that this algorithm effectively solved the serious speech signal distortion problem caused by fixed frequency and white Gaussian noise, and also improved the speech quality.