When classifying a sequence of objects, in an ordinary classification, where objects are assumed to be independently drawn from identical information sources, each object is classified independently. This assumption often causes deterioration in the accuracy of classification. In this paper, we consider a method to classify objects in a sequence by taking account of the context of the sequence. We define this problem as component classification and present a dynamic programming algorithm where a hidden Markov model is used to describe the probability distribution of the object sequences. We show the effectiveness of the component classification experimentally, using musical structure analysis.