METHOD FOR DIVIDING INITIAL STATE BY DIVIDING INTO A SYLLABLES AND A PHONEME, SYSTEM FOR IMPLEMENTING THE SAME

摘要

On the syllable and phoneme in isolated audio recognition method, initially a of the invention is separated, (a), pass through audio signal to a high-pass filter, execute a preprocessing, extracts a scheduled short signal by increasing appropriate window in voice signal; (B) from short signal, across zero ratio, the energy of each framework of the voice signal of the first auto-correlation coefficient, the one or more features variable containing spectral energy ratio is calculated; (C) separation uses the syllable of characterisitic parameter; (D) (c) is separated in a separating step by using characterisitic parameter to phoneme syllable; (e) individual phoneme of distribution (d) initial state in step; a ;According to the present invention, an audio recognition method is provided, by separating phoneme, speed becomes high, and the accuracy of the branch in initial state or even a speech recognition ratio significantly improve it is characterized in that using a variable syllable.