VOICE RECOGNITION METHOD AND METHOD OF TRAINING OF VOICE RECOGNITION APPARATUS

摘要

PURPOSE: To let a user more freely speak by requiring best matching between a specified model and an acoustic event to occur continuously to best matching between a general model and the acoustic event for a limitation step. CONSTITUTION: A feature vector, namely, a linear predictive code(LPC) and a cepstrum 11 are sent to a model matching step 13 but the model matching step 13 includes proper grammatical rules and compares the feature vector of an expression s (n) from an audio signal with two kinds of reference model words. The result finally evaluated best from a determination rule box 14 is transferred as the best keyword, namely, as the keyboard best matched with the feature vector of s (n) by grammar. Thus, the user can more freely speak.