Object-Shape Recognition Based on Haptic Image

This paper introduces a concept of haptic image which contains sufficient haptic data acquired from haptic interaction. Deep mining and proper processing of haptic image may extend the applications to many fields. An approach of haptic shape recognition based on haptic image is presented. Firstly, a glove-like device mounted with pressure sensors and fiber sensors is utilized to acquire haptic image during the exploration of object shape. Secondly, pre-processing of haptic image is conducted including smoothing and standardization. Thirdly, haptic flow is extracted from haptic image as shape feature. Haptic flow proposed in this paper is the displacement of contact points between adjacent time intervals, which is inspired by optical flow. At last, a self-organizing map (SOM) is employed for the classification and recognition of the explored shapes. In the experiment, a recognition test of 4 different shapes, including cube, block, cylinder and sphere, is conducted and the mean recognition rate is approximately 90%.