A speaker recognition system based on VQ

Text-independent speaker recognition is a research hotspot. After 12 MFCCs and centroid of each frame were extracted as audio features of speakers, codebooks of speakers are obtained by LBG-VQ algorithm. At last, speaker identification is done according to Chebyshev distance rather than Euclidean distance of codebooks between testing and each training sequence. The experiments show the proposed system is simple and workable in speaker identification with high speed.