IDENTIFIKASI SUARA MENGGUNAKAN METODE MEL FREQUENCY CEPSTRUM COEFFICIENTS (MFCC) DAN JARINGAN SYARAF TIRUAN BACKPROPAGATION

SPEAKER IDENTIFICATION USING MEL FREQUENCY CEPSTRUM COEFFICIENTS (MFCC) AND ARTIFICIAL NEURAL NETWORKS BACKPROPAGATION METHOD

Abstract

The voice is basic humans tool of communications. Speakers identifications is the process of recoqnizing the identity of a speaker by comparing the inputed voice features with all the features of each speaker in the database.There are two step of speaker identification process: feature extraction and pattern recognition. For the characteristic extraction phase using Mel Frequency Cepstrum Coefficient (MFCC) method. The method of pattern recognition using backpropagation artificial neural networks that compares the test data with the reference data in the database based on the variable result in the learning process.

The result from the research show that increasing SNR (Signal to Noise Ratio) value will determine the success of the speaker recognition system. The higher SNR (Signal to Noise Ratio), will increase percentage level of recognition. Average accuracy speakers recoqnition of the speakers data without noise generating is 86%, the biggest average accuracy speakers recoqnition is  92 % in the data with 80 dB SNR level, and the lowest average accuracy is  45 % in the data with 80 dB SNR level. Rejection rate testing result of speakers outside the database is 100 %.