Home physiotherapy rehabilitation based on RGB-D sensors: a hybrid approach to the joints angular range of motion estimation

Home-based recovery is gradually being used to reduce health-care costs; however, with a shorter stay in the hospital, the risk of growing adverse clinical outcomes exists, mainly due to the lack of motivation in the patient and on the difficulties in performing a strict control by the doctors. This is particularly true for patients who went under knee arthroplasty or total knee replacement who should strictly follow the effective recovery protocols delivered by the doctors. The development of tools for measuring the functional recovery of the operated joint is therefore deemed crucial both for the patient to feel motivated in performing the right number of exercises, and for the doctor that can follow him/her up remotely. One of the most recognized methods for assessing the correctness of a series of recovery exercises, is to monitor the pose of the patient in real-time so as to evaluate its posture in his range of motion. Accordingly, in this paper a novel hybrid approach to 3D human pose estimation is proposed. A first estimation of 2D body pose of the patient in the scene is given, then the depth information coming from the RGB-D sensor is exploited to estimate the joints 3D coordinates. The proposed algorithm proved to overcome the main limitation of using a pure 3D skeleton tracking algorithm during physiotherapy rehabilitation.