Pelvic Trajectory Analysis for Lower Limbs Rehabilitation Robot

Pelvic motion parameters during a gait cycle are important features for gait disorders. The six degree of freedom of pelvic is an important factor to keep balance during a gait cycle. In this paper, we propose a new method to generate a personalized pelvic motion during a gait cycle based on Support Vector Regression (SVR) algorithm. The real-time pelvic displacements can be obtained upon the prototype of lower limbs rehabilitation robot mentioned in this paper. Through the SVR algorithm, we confirm the parameter of training model and generate a pelvic trajectory fitting the data set. We also define a confidence interval around the curve which supposes to be tolerance zone in the coming rehabilitation process or a new subject. Then we use a cross validation method to verify the generalization performance of the curve. The experimental results extracted from our low limb rehabilitation robot demonstrate that the curve generated by us is close to the ablebodied natural pelvis motion. In the end, we discuss the method and consider the future work of our research.