Sistem Pengenalan Suara dengan Metode Mel Frequency Cepstral Coefficients dan Gaussian Mixture Model
ABSTRACT – Biometric technology is becoming a technological trend in various fields of life. Biometric technology utilizes human body parts as a measuring system that is unique for each individual. Voice is a part of the human body that is unique and suitable to be used as a measuring tool in systems that adopt biometric technology. Voice recognition system is one application of biometric technology that focuses on the human voice. Voice recognition system requires feature extraction method and classification method, one of the feature extraction methods is MFCC. MFCC starts from the pre-emphasis stage, frame blocking, windowing, fast fourier transform, mel frequency wrapping and cepstrum. While the classification method uses GMM by calculating the likelihood of similarity between votes. Based on the test results, the MFCC-GMM method in ideal conditions has an accuracy rate of 82.22% while in non-ideal conditions it gets an accuracy of 66.67%.
Keywords – Voice, Recognition, MFCC, GMM, System