Classifying wheelchair propulsion patterns with a wrist mounted accelerometer

In this paper, we describe a manual wheelchair propulsion classification system which recognizes different patterns using a wrist mounted accelerometer. Four distinct propulsion patterns have been identified in a limited user study. This study is the first attempt at classifying wheelchair propulsion patterns using low-fidelity, body-worn sensors. Data was collected using all four propulsion patterns on a variety of surface types. The results of two machine learning algorithms are compared. Accuracies of over 90% were achievable even with a simple classifier such as k-Nearest Neighbor (kNN). Being able to identify current propulsion patterns and provide real-time feedback to novice and expert wheelchair users is potentially useful in preventing future repetitive use injuries.