Complex dynamical analysis of a discrete memristive neural network and its DSP implementation

Abstract

This paper introduces a discrete memristor model and verifies the correctness of the model through circuit simulation. A six-dimensional discrete neural network was built by coupling the Rulkov neuron and the KTZ neuron. Dynamical analyses show that this neural network has multiple firing patterns when the memristor parameters and coupling coefficient are varied in the appropriate ranges, such as periodic firing, quasi-periodic firing, chaotic firing, and hyperchaotic firing. In addition, the coexisting multiple firing patterns and state transition phenomena of this neural network are revealed. Finally, the complexity analysis shows that the generated chaotic sequences have high pseudo-randomness, and the hardware implementation is completed in the Digital Signal Processor (DSP). This paper provides a reference for the study of memristive neural networks and communication encryption.