Aplikasi Identifikasi Nada Darbuka Dengan Onset Detection, MFCC, Dan KNN

Abstract

Darbuka is one of the hadrah musical instruments that acts as a marker when the vocals raise or lower the rhythm of the sound. In learning the Darbuka, the trainer needs to check whether the sound produced is correct or not. With the Darbuka tone recognition system, it will be easier for someone to learn hadrah without a coach. The system developed in this study uses onset detection to break the tone pattern. Then each note goes through a feature extraction process using MFCC with parameters of frame length, overlap length, and the number of coefficients. Then the results of feature extraction through a classification process using KNN. Thexresultsxof the system test showxthat the best combination of parameters in the identification of Darbuka tones with a frame length of 20 ms, overlap length of 40%, the number of MFCC coefficients as much as 13 and a value of K = 1 produces a basic tone identification accuracy of 100%, a tone pattern identification accuracy of 30%, and the accuracy of basic tone identification in the tone pattern is 72.67%.