A SPEECH ENHANCEMENT METHOD USING A SMOOTHED GLOBAL LIKELIHOOD RATIO

摘要

A method for improving the quality of the voice in which the noise is mixed by using SGLR(Smoothed Global Likelihood Ratio) is provided to improve the quality of the voice in which the noises is mixed by modifying a noise elimination gain and evaluate GSAP(Global Speech Absence Probability). The likelihood ratio by each frequency channe is calculated by using equation and GSAP by using SGLR. The noise spectrum is estimated by using GSAP. A noise elimination gain is modified by using GSAP and noise spectrum. By utilizing the noise elimination gain modified with the noise spectrum, the tone quality of the voice in which the noises is mixed is improved.