SYSTEM FOR ROBUST VOICE ACTIVITY DETECTION AND CONTINUOUS SPEECH RECOGNITION IN A NOISY ENVIRONMENT USING REAL-TIME CALLING KEY-WORD RECOGNITION BY CONTINUOUSLY ANALYZING THE RELIABILITY FOR THE SPEECH RECOGNITION

摘要

PURPOSE: A system for robust voice activity detection and continuous speech recognition in a noisy environment using real-time calling key-word recognition is provided to implement continuous speech recognition robust to a noisy environment.;CONSTITUTION: When a speaker says a calling keyword, the calling keyword is recognized by measuring reliability for the calling keyword. The calling keywords from a speaker are recognized by applying a speech section to a continuous speech recognition engine. An adaptive filter removes a surrounding noise by automatically tracking the spectrum properties of the surrounding noises for the voice including the inputted noises. A feature extraction unit performs the conversion of the spectrum features into a robust parameter by modeling the spectrum features for the voice passing through the adaptive filter.;COPYRIGHT KIPO 2010