Low cost RGB-D vision based system for on-line performance evaluation of motor disabilities rehabilitation at home Flavia Benetazzo ∗ and Sabrina Iarlori and Francesco Ferracuti and Andrea Giantomassi and Davide Ortenzi and Alessandro Freddi and Andrea Monteri` u and Marianna Capecci and Maria Gabriella Ceravolo and Silvia Innocenzi and Sauro Longhi Abstract Physical rehabilitation is an important medical activity sector for the recovery of physical functions and clinical treatment of people affected by different patholo- gies: neurodegenerative diseases (i.e. multiple sclerosis, Parkinson and Alzheimer diseases, amyotrophic lateral sclerosis), neuromuscular disorders (i.e. dystrophies, myopathies, amyotrophies and neuropathies), neurovascular disorders/trauma (i.e. stroke and traumatic brain injuries), and mobility for the elderly. During the rehabil- itation, the patient has to perform different exercises, which are specific for the own disease, supervised by physiotherapists. While some exercises have to be performed with specific equipment and under the supervision of professional staff, others can be performed by patients without the supervision of physiotherapists. This allows to reduce the costs of health and care national system and to accomplish the treatment at home. In this work, a computer vision system for physical rehabilitation at home is proposed. The vision system exploits a low cost RGB-D camera and open source libraries for the image processing in order to monitor the exercises performed by the patients, returns a video feedback to improve the treatment effectiveness and increase the user’s motivation, interest, and perseverance. Moreover, the vision sys- tem evaluates an exercise score in order to monitor the rehabilitation progress, this is helpful both for the clinician staff and patients. In this way the physiotherapists can also monitor the patients at home and correct their posture if the exercises are not well performed. This approach has been implemented and experimentally tested using the Microsoft Kinect camera, demonstrating good and reliable performances. ∗ Corresponding author Flavia Benetazzo, Sabrina Iarlori, Francesco Ferracuti, Andrea Giantomassi, Davide Ortenzi, Alessandro Freddi, Andrea Monteri` u, Marianna Capecci, Maria Gabriella Ceravolo, Silvia Inno- cenzi and Sauro Longhi are with: Universit` a Politecnica delle Marche, Via Brecce Bianche, 60131 Ancona, Italy e-mail: f.benetazzo, s.iarlori, f.ferracuti, a.giantomassi, d.ortenzi, a.freddi, a.monteriu, m.capecci, m.g.ceravolo, s.innocenzi, sauro.longhi@univpm.it 1 Revised personal version