Performance analysis of hybrid model of robust automatic continuous speech recognition system

Abstract

In this work, we evaluate the performance of objective measures of noisy input of continuous speech signal through the Hybrid method using Voice Activity Detection (VAD) and Speech Enhancement Algorithm (SEA). Automatic Speech Recognition (ASR) is an important technology, which enables natural human-machine interaction, for over five decades. The objective of this work consists in working out an identification of continuous speech recognition. The methodology presented allows evaluating the process which includes a speech-to-text system using continuous word recognition with a vocabulary of ten words (digits 0 to 9). In the training period, the continuous digits are recorded using 8-bit Pulse Code Modulation (PCM) with a sampling rate of 8 KHz and save as a wave format file using sound recorder software. For a given word in the vocabulary, the system builds an Hidden Markov Model (HMM) model and trains the model during the training phase. The training steps, from VAD, Speech Enhancement to HMM model building, are performed using PC-based Matlab programs. An overall Recognition Accuracy (RA) of 72.45% is achieved from the proposed speech recognition system working under different environment condition for an uttered word.