摘要： 隐马尔可夫模型(Hidden Markov Model) 是一种双随机过程,被广泛地应用于信号处理和模式识别中。但当将其应用于聚类时, 对隐马尔可夫模型的训练，即参数估计，是一个非常重要的问题,训练方法的优劣将对整个应用效果产生重要的影响。本文针对这一问题，将自劈分合并竞争学习运用于HMMs。实验结果证实了本文提出方法的有效性。

Abstract: Hidden Markov Model has been widely used as a valuable stochastic model. A novel clustering method based on both self-splitting and merging competitive learning and HMMs is proposed to improve the quality of the clustering. Experimental results show that our algorithm is very effective for such gene expression datasets.