A novel method for quantifying arm motion similarity

This paper proposes a novel task-independent method for quantifying arm motion similarity that can be applied to any kinematic/dynamic variable of interest. Given two arm motions for the same task, not necessarily with the same completion time, it plots the time-normalized curves against one another and generates four real-valued features. To validate these features we apply them to quantify the relationship between healthy and paretic arm motions of chronic stroke patients. Studying both unimanual and bimanual arm motions of eight chronic stroke patients, we find that inter-arm coupling that tends to synchronize the motions of both arms in bimanual motions, has a stronger effect at task-relevant joints than at task-irrelevant joints. It also revealed that the paretic arm suppresses the shoulder flexion of the non-paretic arm, while the latter encourages the shoulder rotation of the former.