Detection of the start/end of words for word recognition

摘要

A speech recognition process uses an algorithm to identify the start of speech and simultaneously the end of a pause as well as the end of speech. The speech signal is divided into a number of blocks and a characteristic vector is defined that has two characteristics. One is a function of the signal energy and the other is a function of the quadratic difference of a linear predictive coding, LPC, Cepstrum coefficient. The average characteristic vector is calculated based upon a fixed number of blocks and a comparison made with a threshold value to determine start and end points.