Enhancing the tracking capabilities of the Microsoft Kinect for stroke rehabilitation

Abstract

Current motion tracking systems used in games for upper limb stroke rehabilitation either use sensor based tracking systems that can be problematic for a user with an impaired upper extremity to wear or hold, or vision based tracking using cameras that are less robust and struggle to retrieve accurate three dimensional position information. The Microsoft Kinect is a low cost depth sensor that promises to provide a convenient markerless interface for the user, while maintaining the robustness and 3D tracking capabilities of sensor based devices. Evaluation of Kinect's operating characteristics and safety issues is presented. Implementation of a robust hand tracking software system to enhance the tracking capabilities provided with the `out of the box' SDK is also detailed. This system is used as an input method for several stroke rehabilitation games, designed in collaboration with Health Psychologists, Stroke Users Groups, Community Support Team for a previous project based on Nintendo Wii hardware. This study demonstrates that the Kinect-based system presents the same tracking capabilities as our previous virtual glove project using Nintendo Wii hardware, whilst overcoming limitations imposed by the requirement of wearing a physical tracking device.