Array computing based system for visual servoing of neuroprosthesis of upper limbs

Abstract

A hardware-software system implemented for visual servoing of prosthetic arms. The prosthetic system comprises glass worn eye-tracker and video camera and a prosthesis - mounted stereo camera system. The proposed light-weight architecture has to be implemented as a wearable device. We propose a hybrid solution which contains object selection and recognition, in glass camera view, object matching to the prosthesis-mounted camera to compute depth map for object reaching by the prosthetic arm. The object selection and object matching both use the Scale Invariant Feature Transform (SIFT) algorithm. The SIFT algorithm is time-consuming, so the FPGA acceleration is proposed for the light-weight device. The proposed implementation is compatible with real-time servoing of the prosthetic arm.