An efficient approach to human motion recognition employing large motion-database structure

In this paper, an efficient method for human motion recognition is proposed using a large motion-database structure. The database structure is created and managed using a fast-searching structure ldquoB-treerdquo. Different motions of different persons are captured from different directions at the time of learning. Those motions are compressed, projected onto an eigenspace, and stored within the B-tree as indexes. The indexes are technically searched at the time of recognition. For testing the effectiveness and speedy retrieval within the large motion database, the recognition was performed with 60 test motions and satisfactory results were obtained.