Towards rehabilitative e-Health by introducing a new automatic scoring system

The global adoption of consumer devices capable of wide connectivity like smartphones and tablets have led to improvements in their support infrastructure. These make e-health systems for rehabilitation entirely feasible. Current assessments of patient condition can be subjective and inconsistent as the monotony of repetitious tasks lowers alertness. We propose a system to automate the scoring process for the patient's state. This is performed by embedding widely available sensors such as accelerometers sensors into the objects used in a rehabilitative assessment. These sensors introduce signal distortions such as drift and noise which require data driven filtering as the trajectories of human motion are statistically nonstationary. Building on previous work, we compare the use of time and transform domain processing of motion signals by using splines and singular spectrum analysis on the signals and use data analytic techniques for deriving the assessment scores with good results. These form the basis of an e-health system which is evidence-based, and provides the basis for gains in efficiency and a higher level of healthcare.