A tactile sensing surface for artificial neural network based automatic recognition of the contact force position

Abstract

The paper describes a new technique for data entry hardware. It is a high-level interpretation method based on the back-propagation neural network model, in which two aspects of research and development works are illustrated in detail: one is a tactile sensing surface comprising a deformable surface with optical sensing elements; the other is the data acquisition and processing model in which a neural network model, called Tactile Position Recognitron, is programmed to realize the real-time and precise recognition of a contact force position, which enables the contact position of a constant force to be determined within an accuracy less than 4 per cent of full scale in a continuous spatial resolution of 35 zones.

The device described utilizes a simple low-cost and robust mechanical design combined with software to interpret sensory data to measure the contact position of a normal force applied on a planar surface. The high-level interpretation method for this system enables automatic determination of contact position in real time.