Realization of speech recognition using DSP

Abstract

Huge data processing in speech recognition is a tendency. But it is not easy to develop and implement speech recognition system in a small machine. For that reason, the authors have developed a PC-independent speech recognition system using a digital signal processor (DSP). They performed experimental speech recognition tests for some ordinary words. To test the online performance of the system, they developed a program that would detect environmental noise and the endpoint of speech in real time. In experiments, they used 10/sup th/ order cepstrum coefficients and log-scaled energy as the feature vectors and the discrete hidden Markov model (DHMM) as the recognition tool. For the experiment, they selected 5 speeches of 5 males per model for training.