Design and Implementation of Intelligent Singer Recognition System

Abstract

An intelligent singer recognition system was designed to identify the singer. The scheme established a song library at first, then used MATLAB to extract Mel Frequency Cepstral Coefficients (MFCC) from each song in the song library, moreover, set up characteristic parameters pattern base and trained the pattern base by Vector Quantization (VQ) to obtain the final codebook base. Finally, it can correctly classify the singer based on Dynamic Time Warping (DTW) matching reference characteristic parameters pattern with test pattern. Test results showed that the system’s recognition rate is up to 90%.