Quantitative Evaluation of Postural Recovery During Human Walking by System Identification

This paper proposes a method to quantitatively evaluate human posture recovery performance when mechanical disturbances are applied during walking, considering that human falls mainly occur while walking. During walking, the reaction force from the soles of the feet acts as a disturbance during normal walking, which makes it difficult to distinguish between postural responses due to intentional mechanical disturbances and those due to normal walking. Furthermore, it is not easy to reproducibly apply mechanical disturbances during walking. The study uses a treadmill with independent left and right belts. The speed difference between the left and right belts is used as the disturbance input, and the dynamics of human postural correction due to disturbances is identified using system identification with the response of the center of pressure as the output. A method to quantitatively evaluate the posture recovery ability is proposed using the pole distribution of the identified dynamics model. The effectiveness of the proposed method is shown by experimentally demonstrating that there is a quantitatively significant difference in posture recovery performance when posture recovery ability is artificially reduced compared to when it is not reduced.