Model-based clustering is one of the most important ways for time series data mining. However, the process of clustering may encounter several problems. In this paper, a novel clustering algorithm of time-series which incorporates recursive Hidden Markov Model(HMM) training is proposed. Our contributions are as follows: 1) We recursively train models and use these model information in the process agglomerative hierarchical clustering. 2) We built HMM of timeseries clusters to describe clusters. To evaluate the effectiveness of the algorithm, several experiments are conducted on both synthetic data and real world data. The result shows that the proposed approach can achieve better performance in correctness rate than the traditional HMM-based clustering algorithm.