Recognition of complex surfaces based on multiscale temporal networks

Surface recognition plays an important role in many applications as diverse as object detection, robotic grasp, and environmental exploration. However, tactile-based surface recognition is still an open challenge due to the lack of skin-comparable tactile sensing systems and smart pattern recognition algorithms. From the haptic exploration of objects by humans, we propose a novel framework for surface recognition, where the identity of the surface is obtained by the dynamic evolution of the force on the sliding finger. Specifically, we design a flexible film sensor array based on force-sensitive resistors (FSRs) to obtain the change of pressure during finger sliding and obtain the identity of the surface from such spatiotemporal pressure data. We develop a Multiscale Temporal Network (MTCN) to extract discriminative features from pressure data at multiple scales in the temporal domain. Besides, we use a squeeze-and-excitation layer to adaptively calibrate the feature channels from different scales and improve the performance of the model. The experimental results of the six texture classification tasks show that the recognition rate of the proposed method achieves 98%, which validates the effectiveness of the proposed model. We propose a novel way for tactile object surface recognition, which is convenient but powerful.