Artificial vision system design and implementation based on BaSrTiO3 & Nd2O3 composite memristors for efficient pattern recognition

Abstract

In the context of the continuous development of artificial intelligence technology, artificial vision systems play a crucial role as an integral component, mimicking human visual functionality. However, traditional artificial vision systems are typically implemented through algorithms that require substantial computational power during operation. Therefore, finding a more efficient method to emulate biological vision systems remains a significant challenge. To address these challenges, we fabricated memristor devices with barium strontium titanate (BST) and neodymium oxide (Nd₂O₃) composite materials as the functional layer, which exhibit optoelectronic properties. Compared to pure BST devices, BST&Nd₂O₃ devices demonstrate superior electrical performance, stability over 1 × 10⁴ times, and light response with a 4 μA photocurrent that can be repeated. Testing results revealed that the devices possess the capability to simulate synapses and photoreceptor cells simultaneously. Therefore, based on the performance of the devices and the leaky integrate-and-fire (LIF) neuron model, we designed and constructed an artificial vision system. In this system, the BST&Nd₂O₃ device is used to simulate the photosensitive cells of the biological retina and the synapses connected to LIF neurons, while the LIF artificial neurons are constructed using hardware circuits. By leveraging the device's high photoresponse current of 4 μA and its ability to emulate the learning capabilities of biological synapses, we simulated the response to light signals and the ability to recognize digits. The system can successfully recognize 10 different digits after just 10 light signal trainings. Experimental results demonstrate that the constructed artificial vision system exhibits excellent performance and recognition capability, providing new avenues and methods for the application of novel materials in the field of artificial intelligence.