Self-powered high-sensitivity sensory memory actuated by triboelectric sensory receptor for real-time neuromorphic computing

Artificial sensory memory, which is expected to collect, integrate, and refine massive sensory data timely for dynamically training the bioinspired neural network, is a promising candidate to achieve novel architectures of hardware artificial intelligence to mimic neural network. Unfortunately, the reports about artificial sensory memory are very limited and more importantly, there are still many unsolved problems in previously reported artificial sensory memory devices, such as the low sensitivity of perception receptors, high power consumption, and realization of instantaneous neuromorphic computing. Here, we propose a rapid-response, high-sensitivity, and self-powered artificial sensory memory, which is integrated with a triboelectric nanogenerator (TENG) and a field effect synaptic transistor, and is able to achieve real-time neuromorphic computing with a TENG matrix for the first time. Typical properties of sensory memory are successfully demonstrated, such as, excitatory postsynaptic current and paired pulse facilitation, followed with hierarchical memorial processes from sensory memory to short-term memory and to long-term memory. Finally, 28 x 28 matrix triboelectric sensory receptors are fabricated to connect the real-time handwritten image with large-scale data processing. This work proposed a remarkable self-powered artificial afferent nerve to realize rapid and high-sensitivity response, which would show a widespread potential in low consumption artificial neuromorphic interface such as human-robot interaction, edge computing and neurorobotics.