Kinect-based approach to upper limb rehabilitation

Abstract

In this chapter, rehabilitation training is analyzed concerning the possible representation where real-time assessment of correct execution has the highest priority. In order to ensure proper execution of rehabilitation training, real-time feedback to the patients needs to be provided. As a first step in this task, the efficient means to process measured data to generate this feedback will be investigated and realized. Since rehabilitation exercises are sequences of body part movements, there exists a causal knowledge related to exercise movements. The complex relationship between body parts shall be represented by Answer Set Programming (ASP) allowing us to perform causal reasoning. This chapter also discusses the tool used to create individualized training sequences or exercises for stroke rehabilitation patients with respect to design decisions, general architecture, usage, and adjustability. The tool provides an interface to the depth sensor used to capture a patient and evaluates exercises via an interface to an ASP solver for causal reasoning. Additionally, the application offers a graphical user interface (GUI) as well as a general interface for the integration of the tool into the StrokeBack framework. Furthermore, the steps a physical therapist has to perform to create a new training sequence or to create a compensation movement schema in cooperation with a stroke rehabilitation patient are explained and a detailed description on how compensations are handled in ASP is given.