We attempt to make a model based on GMM that describes the feature obtained from the sound signal
of music in relation to the feelings of it. In this paper, we think application the diversity of the impression received
from music to the model. The feeling of music do not only depend on the individual, but also have a common part
in a lot of people. So it is necessary to consider the commonness and the individuality of it. Then, we propose the
technique for appropriately classifying the feeling estimation model responding to the tendency to different feelings.