Adaptive Speech Enhancement Method Using Noise Masker Effect

摘要

PURPOSE: A method for making speech quality better using a noise mask effect is provided to extract noises by masking, thereby preventing musical noises and voice distortion. CONSTITUTION: A noise is estimated from a voice which is contaminated at SNR(Signal to Noise Raito) 10dB by white Gaussian noise. It is estimated that the noise is larger than a real signal. 0 is applied to the noise. The quality of the voice becomes better by a spectrum subtraction method. An NSS(Nonlinear Spectral Subtraction) method based on gain function GiOS(f) is applied to suppress a musical noise. An overweight subtraction element is included in a nonlinear SS method based on gain function GiOS(f).