Ar, İlktanAkgül, Yusuf Sinan2019-06-282019-06-2820137https://hdl.handle.net/20.500.12469/1404https://doi.org/10.1007/978-1-4471-4594-3-50This paper describes a robust low-cost vision based monitoring system for home-based physical therapy exercises (HPTE). Our system contains two different modules. The first module achieves exercise recognition by building representations of motion patterns stance knowledge and object usage information in gray-level and depth video sequences and then combines these representations in a generative Bayesian network. The second module estimates the repetition count in an exercise session by a novel approach. We created a dataset that contains 240 exercise sessions and tested our system on this dataset. At the end we achieved very favourable recognition rates and encouraging results on the estimation of repetition counts. © 2013 Springer-Verlag London.eninfo:eu-repo/semantics/openAccessConference Object48749410.1007/978-1-4471-4594-3-502-s2.0-84887835956N/AN/A