ELM for the Classification of Music Genres
Qi-Jun Benedict Loh, Sabu Emmanuel · 2006
As we produce more digital music, they need to be organized into various classes of music for easy search and retrieval operations. Various classifiers can be employed to carry out the classification. This paper evaluates the performance of extreme learning machine (ELM) as a classifier in the field of music classification. Core components of the classification system include music features, which need to be benchmarked with the ELM. Zero crossing rates, energy, root-mean-square, crest factor, spectral centroid, Mel-frequency cepstral coefficients and specific loudness sensation were features used in this study. We compare the classification accuracy results of ELM classifier against that of support vector machine (SVM) classifier. The classification accuracy results were comparable, with ELM having 85.3125% accuracy and SVM 82.8125%