In order to observe the underwater behavior of sperm whales, the authors previously introduced a passive acoustic system that consists of an AUV (Autonomou Underwater Vehicle) with a hydrophone array and two arrays attached to a support ship. The three arrays receive sperm whale clicks and determine the direction to the sound source by a SBL (Short Baseline) system. Based on triangulation with a pair of arrays, it is possible to calculate the location of the sound source in real-time. Sperm whales usually dive in groups, and so each array captures thousands of clicks from multiple whales. It is, therefore, necessary to identify the corresponding clicks of each individual captured by the three arrays. However, the acoustic communication band between the AUV and the support ship is limited, so it is impossible to correlate the clicks received by each array in real-time. In order to triangulate with a limited communication band, the authors introduce a click classification scheme at each array considering the individual sound source. The class data is sent together with direction data to the AUV and an array on the support ship. Thus the AUV and the operator in the support ship can estimate the posisiton of the whale by a LBL (Long Baseline) system. In this paper, the proposed classification algorithm is applied to data recorded off Ogasawara in 2003. The clicks can be classified reasonably by the proposed method. The three-dimensional underwater trajectories of six sperm whales are obtained by using the classes obtained by the proposed method.