A Robot Hand with Capacitive Tactile Sensor for Object Recognition using Support Vector Machine

Abstract

Touch sensing is essential for humans to perceive objects. Tactile sensing is also important for robot hand manipulation as it provides information on surface properties and interactive forces at contact points between robot fingers and an object. This paper presents a robot hand grasping system equipped with a novel capacitive tactile sensor. The sensor is integrated with a high-precision digital-analog hybrid chip based on a novel R-SpiNNaker architecture. According to the tactile dataset obtained from this system, object recognition is performed by the support vector machine (SVM) algorithm. The results show that superior performance is achieved by SVM with kernel of radial basis function, with accuracy of 98.6%.