Comparison of LPCC and MFCC for isolated Malay speech recognition

Abstract

Automatic speech recognition (ASR) is an area of research which deals with the recognition of speech by machine in several conditions. ASR performs well under restricted conditions (quiet environment), but performance degrades in noisy environments. This paper presents a simple experiment by using famous feature extraction method (LPCC and MFCC) and k-NN classifier with ten-fold cross validation to investigate the sensitivity of Malay speech digits to noise by adding 0dB white Gaussian noise. There are four steps to design and develop the Malay speech digits recognition system. They are Digit syllable structure and Malay speech corpus, end-point detection processing, feature extraction and classification method. The average recognition rates for Malays digits recognition is 95% that the feature vectors were derived from LPCC and MFCC in clean environment. The objective of this paper is to shown the occurrence of noise during Malay speech recognition with different features vector (LPCC and MFCC) and comparison between them in noisy environment