Pengaruh Metode MFCC Dan KNN Pada Music Information Retrieval Terhadap Klasifikasi Genre Musik

Abstrak

Music is an important part of daily life, but the vast array of choices makes selecting songs challenging. This research aims to assess the accuracy of music genre classification using Mel Frequency Cepstral Coefficients (MFCC) and K-Nearest Neighbor (KNN) for five popular genres in Indonesia: pop, rock, dangdut, hip-hop, and jazz. The training data consists of 500 samples (100 per genre). The results indicate that the audio cut point in the middle, with a duration of 20 seconds, using the KNN classification method achieves the highest accuracy with a value of k=7, yielding the best accuracy of 68.67%. The middle part of the audio is more representative and informative for genre classification. Therefore, it is recommended to use a middle cut of 20 seconds with k=7 for more accurate classification using MFCC and KNN.