International Journal of Engineering and Technology

Speaker recognition is one of the most essential tasks in the signal processing which identifies a person from characteristics of voices. In this paper we accomplish speaker recognition using Mel-frequency Cepstral Coefficient (MFCC) with Weighted Vector Quantization algorithm. By using MFCC, the feature extraction process is carried out. It is one of the nonlinear cepstral coefficient functions. Then the pattern matching is accomplished by evaluating the similarity of the unknown speaker and the trained models from the database. For this process, weighted vector quantization is proposed that takes into account the correlations between the known models in the database. Experimentations express that the new methodologies provide higher accuracy and it can observe the correct speaker even from shorter speech samples more reliably.