Indian Journal of Science and Technology

Abstract

Objectives: To provide detailed study of all the algorithms and analyse them in a way to conclude that an integration approach describes the best combination for voice recognition in terms of recognition rate. Methods: Voice processing is a process in which words of a speaker are recognized by the information of the waves. There are number of algorithms used for voice recognition named as Perceptual Linear Prediction, Linear Predictive Code, Mel Frequency Cepstral Coefficient, Dynamic Time Warping etc. Findings: Graph is used to depict the recognition rate of all the voice recognition techniques with different types of classifiers named as HMM (Hidden Markov Model), DTW ( Dynamic Time Wrapping), VQ (Vector Quantization), SVM (Support Vector Machine) etc. which clearly describes that the hybrid approach may provide better results as compared to individual methods. Performance and recognition rate is not so good by using individual techniques because it does not provide better recognition rate while taking into consideration the security of an individual living alone at home. After the comparative analysis, it is concluded that there is a need to develop a better combined approach which may provide better recognition rate as compared to individual methods. This paper will help the researchers to know the basic difference between the explained feature extraction techniques. Application: Main application we are using is Voice Recognition in order to provide security to an individual living alone at home.

Keywords: Feature Extraction, LPC, MFCC, PLP, Voice Recognition