ONLINE MODEL VARIABLE COMPENSATING METHOD FOR VOICE RECOGNITION AND VOICE RECOGNIZING METHOD IN ACCORDANCE WITH COMPENSATING METHOD, ESPECIALLY IN RELATION TO COMPENSATING FOR MODEL VARIABLES USED FOR VOICE RECOGNITION ACCORDING TO ENVIRONMENT CHANGE WITH INPUTTED VOICE SIGNALS ONLY WITHOUT ANOTHER ADAPTIVE DATA

摘要

PURPOSE: An online model variable compensating method for voice recognition and a voice recognizing method in accordance with the compensating method are provided to compensate for model variables used for voice recognition according to environment change with inputted voice signals only without another adaptive data, thereby quickening a voice recognition calculation time while easily realizing the time. CONSTITUTION: A feature of an inputted voice is extracted(400). Model variables necessary for voice recognition are compensated for in consideration of changes of noise environment of an inputted voice, and the voice is recognized by using the compensated model variables(410). Voice recognition results are compared with past recognition results, to confirm whether the compared results are the same(420). If not, a noise feature is reestimated(430). In consideration of the reestimated noise feature, the model variables are compensated for(440).