After calculating the points (x,y), these points are mapped onto Thayer’s model to determine emotion [2, 3]. For music emotion visualization, emotion should be detected in each short-time segment. The length of a segment is often less than or equal to 16 seconds such as one second [1], ten seconds and 1/3 overlap [9]. However, there are some weaknesses when using Thayer’s model to recognize the music emotion in each short-time segment. The points near the center of Thayer’s model have different emotions despite their distances are very close (small change of acoustic features) such as A1, A2, A3 in figure 2. Conversely, the distance between A2 and A4 is so far but they have the same
