Study on Similarity Measure of Human Body Motion

身体動作の類似度に関する研究

2020-09-24

2020-09-24

The study introduces non-model-based approaches for measuring the similarity of human body motions through measuring the dissimilarity of marker trajectories in human body motions. The proposed methods in the research are independent of the biological characteristics of the human body. In general terms, the human body motion generates three-dimensional (3D) curves representing the direction changes of marker trajectories. The first or original method is based on the dissimilarity measure for 3D curves. In detail, the dissimilarity of human body motions is estimated by applying the dissimilarity measure of the 3D curves for the dissimilarity of 3D curves representing the trajectories of markers and is implemented in five steps. Firstly, the motion speed is adjusted. Next, human postures in motion are normalized. Then, motion paths of markers are represented by direction sequences. After that, chain codes of direction sequences are obtained. Finally, the dissimilarity of human body movements is calculated based on the obtained chain codes. The roles of markers in motion are also taken into account. In addition, two options of the spatial quantization method for the human active space in a body motion based on the Space Harmony Theory is also mentioned. In order to reduce the conducted time of the original method, the extended method to calculate the dissimilarity of human body motions through sub-motions is also proposed. In the extended method, two options of the pause derivation method for a human body motion are introduced. Human body motions are split into sub-motions by using the pause waveforms obtaining from the pause derivation method. Because the sub-motions are considered as new motions, the dissimilarity of sub-motions can be calculated by using the original method. Both the original method and the extended method are experimentally evaluated by using the statistical test. Experimental results proved that both two methods are effective in the estimation of dissimilarity of the human body motions with the appropriate Split value. The result shows that the extended method improved the conducted time and can be used in real-time applications. Furthermore, the first option of the extended method is effective in enhancing the performing time. The proposed methods are proved effective in measuring the similarity of human body motions.

博士(工学)

京都工芸繊維大学

14303甲第974号

http://hdl.handle.net/10212/2603

eng

ETD

Thesis or Dissertation

Thesis