A novel memristor-based bionic neural network circuit with crossmodal integration and forgetting effects

The development of brain-inspired artificial intelligence is based on an understanding of brain cognitive functions, which are influenced by cross-modal interactions between learning and memory. In this paper, inspired by the neural mechanisms and biological phenomena of visual-olfactory associative memory, a cognition driven bionic cross-modal associative memristor neural network (CAMNN) is proposed. This network includes the complete processes of learning, memory, and forgetting to simulate the interactions of cross-modal associative memory in the brain. The designed circuit primarily consists of three modules: (1) The voltage control module implements sensory information input and integration. (2) Synaptic Module: This module utilizes memristors to simulate synaptic weights for associative memory. (3) The delay module implements memory storage in the brain, thereby maintaining long-term responses. Based on different trigger conditions, the proposed circuit achieves innovative functions such as threshold detection, unimodal learning, cross-modal learning, and cross-modal reinforcement learning. The simulation results in PSPICE demonstrate that the proposed circuit can simulate interactions between different modalities in the brain from multiple perspectives. It also provides a reference for the development and utilization of intelligent robots and neural network circuits.