Speaker Identification Using MEL Frequency Cepstral Coefficients and Vector Quatization

Abstract— In this paper, we build a VQ-based speaker identification system. The speaker identification, which consists of mapping a speech signal from an unknown speaker to a database of known speakers, i.e. the system has been trained with a number of speakers which the system can recognize. Here developed, Text-dependent systems require the speaker to utter a phrase like digits zero to nine in an isolated way. Speaker identification has been done successfully using Vector Quantization (VQ). This technique consists of extracting a small number of representative feature vectors as an efficient means of characterizing the speaker specific features. Using training data these features are clustered to form a speaker-specific codebook. In the recognition stage, the test data is compared to the codebook of each reference speaker and a measure of the difference is used to make the recognition decision. The paper shows identification rate when triangular, or rectangular or hamming window as well as codebook size increases, the identification rate for each of the three cases increases.