Application of Supervised Principal Motion Analysis to Evaluate Subjectively Easy Sit-to-Stand Motion of Healthy People

Abstract

Redundant human motions such as walking or sit-to-stand motions involve time-series data of several variables. Principal motion analysis (PMA) can be adopted to decompose such motions into independent motions, and their linear combinations can be used to approximate the motions. In contrast to the existing PMA methods, which are unsupervised, we applied partial least-squares regression to perform PMA such that the scores for the principal motions were correlated with a continuous objective variable. To validate the practicality of this approach, we investigated the subjectively easy sit-to-stand movement of healthy people. The participants were six healthy young individuals who performed the sit-to-stand movement under 33 different conditions by changing the foot position, hand-grip position, and initial pitch angle of the upper body. The motion data and magnitude of the subjective burden reported for each movement were analyzed. Three principal motions correlated with the subjective burdens were determined and interpreted. The correlation coefficients of the first, second, and third principal motions and the subjective burdens were 0.60, 0.27, and 0.19, respectively. Moreover, the sit-to-stand conditions synthesized by the three principal motions incurred a burden subjectively smaller than or comparable to the burdens in other conditions.

Keywords

• Motion synergy
• foot position
• handrail grip position
• subjective burden
• supervised learning