Tan, Steven Kwang San

Non-proficient demonstration, gross motor skill assessment, and subjective feedback are but a few of the perennial problems in physical education (PE). These problems stand to benefit from a technology-based solution that uses human pose estimation to guide learning. In this approach, a criterion motor action is embedded in a deep-learning algorithm (DLA). A learner can view this motor action on an iPad and uses its kinematic signatures to guide practice. The learner’s movement is captured by the device and the recorded motor action enters the DLA for computation of movement proficiency. The output of the DLA is a quantitative index that informs the learner how well the movement has been executed. In this way, the learner gains timely and objective feedback. A separate device held by the PE teacher collates the quantitative indices from other students in the class. Collectively, the information facilitates the teacher’s selection of instructional strategies.