Identifikasi Pembicara dengan Menggunakan Mel Frequency Cepstral Coefficient (MFCC) dan Self Organizing Map (SOM)

Inggih Permana (Universitas Sultan Sarif Kasim Riau)
Benny Sukma Negara (Universitas Sultan Sarif Kasim Riau)

Seminar Nasional Teknologi Informasi Komunikasi dan Industri
2011: SNTIKI 3

This study tested the accuracy of speaker identification by using the MFCC (Mel Frequency Cepstral Coefficient) and SOM (Self Organizing Map). MFCC is used as feature extraction of voice signal. SOM is used to clustering the feature vectors that obtained from the MFCC. Weight vectors that obtained from the clustering will be used as a codebook. Voice testing is measured the proximity from the codebook by using the Euclidean Distance. The smallest value of the measurement results is the winner vector that representing a particular person. Test results show the SOM can be used for the identification of the speaker but with the highest average accuracy for female speaker only 54.4%, male speaker 75.6% and for all speaker 62.5%. Keywords:  Codebook, Euclidean Distance,  MFCC, SOM, Speaker Identification