APPARATUS AND METHOD SPEECH RECOGNITION BASED INITIAL SOUND

摘要

PURPOSE: An apparatus and method for recognizing speech based on an initial sound are provided to output a result according to the importance of a word searched in an initial sound based speech recognition. CONSTITUTION: A feature analyzing unit generates an input pattern by calculating a feature vector of a speech signal(S20), and the input pattern is compared with a reference pattern(S30). A pattern recognizing unit outputs initial sound information corresponding to the reference pattern similar to the input pattern. A word corresponding to the initial sound information detected in the pattern recognizing unit is searched(S70). According to the searching result, a determination unit determines the word which is supposed to be outputted.