Practical calibation for upper extremety patients in haptic rehabilitation

As haptic devices become more affordable and economical, research in Haptic-based virtual rehabilitation systems is gaining more interest. A prototype haptic virtual rehabilitation system was developed in our lab using the CyberGrasp system with the goal of assisting patients who suffered from stroke leading to temporary upper extremity disability. However, calibration of the system for each user is currently tedious and time-consuming. In this paper, we propose a practical method to calibrate the CyberGlove using a Neural Network. The neural network is trained with the hand segment size and the manually collected calibrated data of 12 subjects. The hand segment sizes are extracted from a 2-D hand image. The preliminary results are very encouraging and show its effectiveness for the sensors on the index, the middle, the ring, the pinky fingers.