Energy-Efficient Neuromorphic Perception

Neuromorphic sensors that leverage the precise timing of neural spikes are opening new avenues for energy-efficient and adaptive artificial perception. By incorporating memristorbased non-volatile memory, these systems achieve high-fidelity processing while maintaining low power consumption, making them well-suited for real-time visual and tactile interpretation. This work introduces a unified visual-tactile processing framework that employs spiking neural networks to integrate multisensory information. The approach enables rapid and accurate perception in robotic applications, including object recognition and slip detection, demonstrating superior performance compared with traditional deep learning methods. Additionally, we provide publicly available visual-tactile datasets to support reproducibility and future research in this area. Our study highlights the potential of memristor-based neuromorphic devices in creating intelligent, low-power robotic systems and discusses their implications for advancing artificial vision and multi-modal sensory processing.