Abstract

This paper proposed a structure optimization principle according to the entropy calculating method of the HMM, and then introduced the TOM as the evaluation stander of the clustering result firstly. Then this paper proposed the WTOM by weighted the TOM, and divided the samples according to the value WTOM. Based on this work, this paper proposed an HMM based time-series images clustering algorithm which named HTICA, HTICA clustered the images hierarchically, theoretical analysis and experiments show that the proposed algorithm is very reliable, it can segment and identify the continuous time series samples unsupervised. And the application results in typical action extraction for badminton athlete is very well.

Keywords: Entropy, Hidden Markov Model, hierarchical clustering, transition occurring matrix, time series distribution.

Graphical Abstract