PROCEDURE FOR PROCESSING NOISY SPEECH SIGNALS, AND AN APPARATUS AND A PROGRAM FOR THE SAME, PARTICULARLY FOR IMPROVING THE SOUND QUALITY USING AN ESTIMATED NOISE

摘要

PURPOSE: A procedure for processing noisy speech signals, and an apparatus and a program for the same are provided to not require a high quantity of memory and reduce the degree of calculation remarkably compared with an existing method.;CONSTITUTION: A procedure for processing noisy speech signals comprises the steps of: obtaining a conversion spectrum by converting an inputted noisy voice signal into a frequency domain; obtaining a smooth magnitude spectrum in which magnitude difference of the converted spectrum is relived between adjacent frames; obtaining a search spectrum that indicates the estimated noisy components of the smooth magnitude spectrum; and obtaining the ratio that indicates how much noises are included in the inputted noisy voice signal by using the search spectrum and the smooth magnitude spectrum.;COPYRIGHT KIPO 2010