Spiking Neural Network Based on Threshold Encoding For Texture Recognition

This paper presents a neuromorphic computing model that classifies material textures using a neural coding scheme based on threshold encoding. The proposed threshold encoding converts raw tactile data of each texture into an event-based data highlighting the spatio-temporal features needed to recognize human touch. Achieved results show that the model can categorize the input tactile signals into their corresponding material textures with high accuracy and fast inference. This work paves the way toward employing the proposed encoding method in more complex tactile based applications from the theoretical and hardware implementation aspects.