Music Information Retrieval Based on Active Frequency

Music is the art of combining frequencies. A balance of frequencies gives rise to a harmonious tone. Several features of music can be analyzed, and they include sociocultural background, lyrics, mood, tempo, rhythm, harmony, melody, timbre, and instrumentation. In this study, we use the frequency of instrumentation as a feature for classification because each instrument has a frequency range. To test this frequency range, we use five music genres and one music playing skill. The five genres are dangdut, electronic dance music (EDM), metal, pop/rock, and reggae. The music playing skill is acoustic. Active frequencies are tested using the k-nearest neighbor method, and the results serve as basis of the accuracy of music classification. The classification accuracy for EDM, metal, and acoustic is over 70%, whereas that for dangdut, pop/rock, and reggae is less than 60%. In sum, the accuracy of music classification is influenced by the similarities in the music instruments used and the tempo.