Fast Speaker Recognition using Efficient Feature Extraction Technique

Digital processing of speech signal and speaker recognitionalgorithm is very important for fast and accurate automatic voicerecognition technology. A direct analysis of the voice signal iscomplex due to too much information contained in the signal.Therefore the digital signal processes such as Feature Extractionand Feature Matching are introduced to represent the voicesignal. The non-parametric method for modeling the humanauditory perception system, Mel Frequency Cepstral Coefficients(MFCCs) is utilized as extraction technique. MFCC imitates thehuman hearing system; therefore it provides better recognitionrates than Linear Predictive Coefficients (LPC). For the presentwork, work, the non linear sequence alignment known asDynamic Time Warping (DTW) is used as features matchingtechnique. Since voice signal tends to have different temporalrate, the alignment is important to produce better performance.This paper presents the viability of MFCC to extract features ofspeech signal and DTW to compare the corresponding testpatterns.