Enhanced MFCC Algorithm using Lookup Table and Kaiser Window

Abstract

Automatic Speaker Recognition System has been quite fascinating for a man from quite long time. There are various feature extraction algorithm like Linear Predictive Coefficient (LPC), Mel- frequency Cepstrum Coefficient (MFCC). In this thesis, enhanced MFCC algorithm has been proposed which reduces the total time by almost 50 percent but accuracy decreases as compared to conventional algorithm from 94 percent to 94.93 percent. But this makes new algorithm to be implemented more effectively on hardware. The proposed algorithm tries to explore in security where speaker recognition can be used for speaker identification. This algorithm uses lookup table and Kaiser Window instead of Hamming Window algorithm as in conventional MFCC algorithm which improves the accuracy of algorithm and lookup table which reduces the time complexity of algorithm also the formula for pre-emphasis has been modified which again reduces the copulation time for pre-emphasis of signal.