Spatio-Temporal Encoding Improves Neuromorphic Tactile Texture Classification

With the increase in interest deployment of robots unstructured environments to work alongside humans, development human-like sense touch for becomes important. In this work, we implement a multi-channel neuromorphic tactile system that encodes contact events as discrete spike mimic behavior slow adapting mechanoreceptors. We study impact information pooling across artificial mechanoreceptors on classification performance spatially non-uniform naturalistic textures. encoded spatio-temporal activation patterns through gray-level co-occurrence matrix computed from time-varying mean spiking rate-based response volume. found approach greatly improved texture comparison use individual mechanoreceptor alone. addition, was also more robust changes sliding velocity. The importance exploiting precise spatial and temporal correlations between sensory channels is evident fact either removal or altering structure pattern, significant drop observed. This thus demonstrates superiority population coding approaches can exploit populations. It, therefore, makes an advance direction bio-inspired systems required realistic applications robotics prostheses.