Abstract

This paper develops computational models to monitor patients with hip replacement surgery. The Kinect camera is used to capture the movements of patients who are performing rehabilitation exercises with both lower limbs, specifically, ‘side step’ and ‘knee lift’ with each leg. The information is measured at 25 body points with their respective coordinates. Features selection algorithms are applied to the 75 attributes of the initial and final position vector of each rehab exercise. Different classification techniques have been tested and Bayesian networks, supervised classifier system and genetic algorithm with neural network have been selected and jointly applied to identify the correct and incorrect movements during the execution of the rehabilitation exercises. Besides, prediction models of the evolution of a patient are developed based on the average values of some motion related variables. These models can help to fasten the recovery of these patients.