The human gait is an important health indicator. Gradual changes to gate are indicators to diseases such as Parkinson’s disease, diabetes, Alzheimer and many others. Yet most technologies existing today for tracking changes in gate are either very expensive or intrusive. In this work we demonstrate how the Kinect sensor can be used to track gait. We build a layer of machine learned models that predict rich set of gait features from the skeleton model extracted by the Kinect SDK. These predictions were calibrated against wearable sensors and were shown to provide accurate predictions.
This approach has several advantages over existing techniques: it is affordable, accurate and not intrusive. Therefore, it allows for continuous monitoring of gait on large populations.