Characteristics in human motion - from acquisition to analysis

摘要

Abstract. The technological progress within the last few years allows the construction of robots that could be part of our everyday live in near future. Besides human like appearance and communication they require natural patterns of behavior and motion to achieve wide acceptance. To analyze human motions we present different kinds of motion capturing systems and post-processing methods for motion data. Further on, we describe principles for the extraction of suitable motion features, especially on the basis of Wavelet Transforms. These features allow the distinction of inter and intra-individual characteristics. Finally, we discuss the design and the application of Hidden Markov Models towards the classification of motion types and the recognition of motion phases. All methods and results are demonstrated by examples of walking and object manipulation motions of different persons and manners. 1