A Study on Classification of Music Pieces by using Self-Organizing Map with Spectrogram

Abstract

In this research, we try to classify music pieces by Self-Organizing Map from the performance sounds represented as images by using spectrogram analysis. By extracting image features from the spectrogram, music can be represented as a set of image features. As a result of the experiment, it is suggested that it is possible to classify the music pieces that reflects the impression of listening it.