Pembangunan Metode Codebook untuk Identifikasi Chord Gitar dengan Teknik Ekstraksi Ciri MFCC

Abstract

Various types of chords can be produced by a guitar instrument. However, it is difficult to recognize a guitar chord manually. This research develops a system to recognize a chord on the guitar using codebook, MFCC, and K-means clustering. The codebook is used as the pattern recognition method for identification of guitar sound, while MFCC is used as a method of feature extraction. The parameters used in the process of MFCC are the number of cepstral coefficients, overlap, and the time frame, while the parameter used in the process of K-means clustering is the number of cluster. This research used 1440 guitar sound data that are divided into 72 class, so that each class has 20 guitar sound data. The results showed that the maximum accuracy obtained is 99.72% when the number of cepstral coefficients are 26, overlap value is 0.4, time frame is 30 ms, and the number of cluster is 5.