Study of feature values for subjective classification of music

Abstract

In this research, we analyze how the sound and music relate to humans from the aspect of Kansei engineering. We analyze what features of the sound humans pay attention and how humans interpret sound. Therefore, we divide the signal processing of sound that humans do into four levels. At the physiological level, processing is done by the auditory characteristic. In this level, humans don't interpret the image of the sound yet. There is no subjectivity for the sound. By using auditory characteristic, we investigate the features which help in the case that sound and music is analyzed. We consider that the processing at early stage of auditory nervous system is to extract the change in power, which is obtained from the segmentation of the sound-signals which is divided by band of the frequency and time interval, and its contrast. We also consider the features obtained by that extractation. Moreover, in the cognitive level, we analyze the correlation of that features with the word of interpretation that humans do subjectively. By these modelings, we develop the method of retrieving the sound and music that having the similarity, or having the image that is expressed by any subjective words. © 2009 Springer Berlin Heidelberg.