Noise reduction of speech signal based on phase spectrum estimation

In the process of short-time Fourier transform (STFT), signal is departed into two components, which are magnitude spectrum and phase spectrum. It was believed that the magnitude spectrum includes most of the information of the speech, and phase spectrum was kept unmodified due to the human auditory system is phase deaf. In this paper, we focused on phase spectrum estimation of noisy speech for a speech-enhancement system. A novel phase spectrum estimation method for speech enhancement system using high resolution STFT is proposed. An approximate value of the phase spectrum is estimated from the high resolution STFT. Then proposed phase estimation method is compared with analysis modification synthesis (AMS) and phase spectrum compensation (PSC) methods. The experiment results show the effectiveness of the phase spectrum estimation and significant contribution to the speech quality.