부산대학교 도서관

Artificial sensory memory is a novel way to solve the contradiction between the information explosion in the era of the Internet of Things (IoT) and the high demand for hardware resources by artificial intelligence, which achieves the integration of sensory and memory by mimicking biological neural system. Here, we innovatively propose an artificial infrared neural system (AINS) based on single-crystal thin films, consisting of artificial receptors, artificial afferent fibers, and artificial synapses. It is implemented by a pyroelectric sensor based on lithium tantalate (LT) thin film, a threshold-based signal processing module, and a memristor based on lithium niobate (LN). After demonstrating the detection ability of the pyroelectric sensor and the short-term plasticity (STP) and time coding ability of the memristor, we successfully coupled them together to achieve the conversion of pyroelectric current (PEC) signal to postsynaptic current (PSC). Based on the above characteristics, the output of an AINS-based visual array for multiple dynamic hand-waving action was further simulated. The results showed that our AINS can achieve intra-recognition of historical events via spatiotemporal fusion imaging. The proposed AINS realizes sensing, memorizing, and processing of sensor information in the analog domain, opening a novel avenue for sensor signal processing.