Kansei Transition Analysis by Time-series Change of Media Content

Abstract

In this paper, we present a new concept, a waveform model of Kansei transition for time-series media content. It is important to apply the time-series change of media content to Kansei information processing. For example, the impression of music media content changes over time. In our model, we represent Kansei transition by time-series change of media content as waveforms. We realize new Kansei similarity by comparison with Kansei transitions represented by waveforms applying a signal processing technique. Through new Kansei similarity, it is possible to realize media content retrieval and recommendation systems corresponding to the time-series Kansei transition of media content. Our model consists of two modules: a high-order media-Kansei transformation module and a waveform similarity computation module. The high-order media-Kansei transformation module extracts each Kansei magnitude by each time from the features of media content. The waveform similarity computation module computes similarities between each waveform represented as Kansei transition.