Augmented-Reality Environment for Locomotor Training in Children with Neurological Injuries

In this paper a novel augmented-reality environment is presented for enhancing locomotor training. The main goal of this environment is to excite kids for walking and hence facilitate their locomotor therapy and at the same time provide the therapist with a quantitative framework for monitoring and evaluating the progress of the therapy. This paper focuses on the quantitative part of our framework, which uses a depth camera to capture the patient’s body motion. More specifically, we present a model-free graph-based segmentation algorithm that detects the regions of the arms and legs in the depth frames. Then, we analyze their motion patterns in real-time by extracting various features such as the pace, length of stride, symmetry of walking pattern, and arm-leg synchronization. Several experimental results are presented that demonstrate the efficacy and robustness of the proposed methods.