Rock Genre Classification using K-Nearest Neighbor

Abstract

Music genre  classification is a part of Music Information Retrieval. This research was a genre music detection based on signal from an audio. Divided into two processes namely extraction of features and classification. Signal would be transformed using Fast Fourier Transform to get frequency domain signal which will be processed to extract Short Time Energy, Spectral Centroid, Spectral Roll-Off, Spectral Flux, and Energy Entropy feature. Besides those features, Zero Crossing Rate would be counted from time-domain signal. in classifying phase, research using k nearest neighbor with accuracy reaching 54,44%.