Abstract Views :210 |
PDF Views:2
Authors
Affiliations
1 Department of ECE, KL University, Vijayawada, A.P, IN
Source
Digital Signal Processing, Vol 4, No 4 (2012), Pagination: 135-141
Abstract
Voice recognition is a process of recognizing a person on the basis of their speech sample. This paper describes various techniques that are used for voice recognition in order to optimize the recognition rate. The different techniques that are described in this paper are Linear Predictive Coding (LPC), Neural Networks (NN), Mel Frequency Cepstrum Coefficients (MFCC), Vector quantization (VQ), Euclidean Distance. MFCC and LPC are used to extract speaker specific characteristics from voice signal. Neural Networks and Euclidean Distance are used for matching the characteristics extracted using MFCC and LPC. The recognition rates are calculated in each method and they are compared. Mel Frequency Cepstrum Coefficients gives better recognition rate when compared with the other two techniques. Various other approaches for implementing voice recognition are Hidden Markov Modeling (HMM), Gaussian Mixture Modeling (GMM), and Dynamic Time Warping etc. The Voice Recognition system has potential applications in various fields. Some of them are access control to computers, telephone banking, forensics, speech recognition etc.
Keywords
Linear Predictive Coding (LPC), Mel Frequency Cepstral Coefficients (MFCC), Neural Networks( NN), Vector Quantization (VQ).