Klasifikasi Genre Musik Menggunakan Learning Vector Quantization (LVQ) dan Self Organizing Map (SOM)

Putri, Luh Arida Ayu Ruhning, Sri Hartati

2016 | Disertasi | FMIPA

Easiness in obtaining digital music files can cause problems in its management. Musical genre classification can help providing genre label of music files, so that management and search of music files can be simplified. The main problem in musical genre classification is to find the combination of features and classifier that can provide the best result in classifying music files into their music genre. This research classifying music files using Learning Vector Quantization (LVQ) that combined with Self Organizing Map (SOM) based on feature of entropy of wavelet coefficients. The combination lies in the initialization of reference vectors of the LVQ which is determined based on the result of clustering the training data using SOM. This is expected could reduce the sensitivity of the reference vector selected directly from training data. The results showed that musical genre classification using a combination of LVQ and SOM gives better results than using LVQ alone, but the accuracy is still low, i.e. 54.23%. Entropy features can not accurately classify 10 genres used in this research. This was shown when classification were performed using the same feature but with different classifiers, the results were also low.

Kata Kunci : musical genre classification; Learning Vector Quantization (LVQ); Self Organizing Map (SOM); entropy of wavelet coefficients feature