The Cognitive Motivation-based APBMR Algorithm in Physical Rehabilitation

Abstract

This article presents a new, alternative method of gesture recognition, using the cognitive properties of intelligent decision-making systems, to support the rehabilitation process of people with disabilities: the Asynchronous Prediction-Based Movement Recognition (APBMR) algorithm. The algorithm “predicts” the next movement of the user by evaluating the previous three with the goal to maintain motivation. Based on the prediction, it creates acceptance domains and decides whether the next user-input gesture can be considered the same movement. For this, the APBMR algorithm uses six mean techniques: the Arithmetic, Geometric, Harmonic, Contrahamonic, Quadratic and the Cubic ones. The purpose of this article besides presenting this new method is to evaluate which mean technique to use with the three different acceptance domains. We evaluated the algorithm in real-time, using a general and an advanced computer, as well as testing, verified by prediction, from a file and comparison of the algorithm to one of their earlier works. The tests were done in four groups of users, respectively, each group performing four gestures. After analyzing the results, we concluded that the Contraharmonic mean technique gives the best average gesture acceptance rates, in the ±0.05 m and ±0.1 m acceptance domains, while the Arithmetic mean technique provides the best average gesture acceptance rate in the ±0.15 m acceptance domain, when using the APBMR algorithm.