Penerapan SOM untuk Pengenalan Nada pada Angklung Modern

Abstract

Angklung is an Indonesian cultural heritage from West Java and has been developed following the development of modern music. The purpose of this research is to recognize the tone of angklung by using self-organizing maps (SOM) and mel-frequency cepstrum coefficients (MFCC) feature extraction. The used data were 300 tones of angklung that were played by staccato of 12 diatonic tones, each tone by 25 voices. Tones data were recorded for 3 seconds with some staccato in play. The initial process was making the data into one staccato tone, normalized, and then its characteristics were extracted with 26 cepstral coefficients MFCC. Distribution of training and test data was conducted randomly by the composition of 19 training data and 5 test data for each tone. This research conducted three repetitions on radius 0, 1, and 2 with training data, test data and learning rate of 10 parameter values at different SOM in each of their repetition. The result shows that the best accuracy obtained was 98.33% at a radius of 1 in repetition 1 and 3, as well as at a radius of 2 in repetition 1 with a common learning rate in both radius, 0.0090