Object Recognition based on Surface Detection - A Review

Abstract

The task of object recognition for prosthetic hands to perform effective grasping has been among the prime objectives in the domain of rehabilitation robotics. Pertaining to the sensors being implemented in the existing approaches, the object recognition techniques can be divided into vision based and haptic based categories. Most of the works on object recognition for prosthetic hands have adapted hybrid approaches by utilizing both vision and tactile sensors together. Nevertheless, classification of visual features along with tactile features incurred high algorithmic complexity and excess hardware requirements. Moreover, embedding a vision based system for a prosthetic limb does not seem to be natural and practically plausible solution. Therefore, an approach towards tactile based object recognition with advantages of the reported hybrid system seems assuring. In this review, existing approaches towards tactile sensing for prosthetic hands are identified. A comparative study and analysis on usability of tactile sensing methods is also considered for discussion in this paper. Based on this review, knowledge representational approaches for object recognition by prosthetic hands are suggested as one of the prominent future directions of research.