Performance Enhancement of Automatic Speech Recognition System using Euclidean Distance Comparison and Artificial Neural Network

The paper shows how an Automatic Speech Recognition (ASR) system be efficiently designed for ubiquitous control. The design is based on an algorithm to extract isolated words from a continuous speech signal. The feature extraction of the voiced part of speech signal is done byMel Frequency CepstralCoefficients (MFCC) whereas Artificial Neural Network (ANN) is used for training and pattern. The increased rejection of unauthorized speech commands is obtained by the decisions based on Euclidean distance, measured between trained and tested voiced commands. SNR of the signal is also improved,at the pre-processing stage.