Fractional-order heterogeneous neuron network based on coupled locally-active memristors and its application in image encryption and hiding

Abstract

Synaptic crosstalk significantly influences neural firing in the brain. Locally-active memristors can effectively emulate neural network synapses and have a significant importance in neural network research. This paper designs a tristable locally-active memristive model and presents a novel fractional-order (FO) heterogeneous neuron network. This neural network consists of Hindmarsh-Rose (HR) neuron and FitzHugh-Nagumo (FHN) neuron, which are connected by coupling FO locally-active memristors. The research found that changes in the order of different dimensions have a significant effect on the neural network firing through the three-parameter bifurcation diagram. Moreover, it is found that the locally-active memristor as a synapse can affect the coexistence firing behavior of the network. The complex dynamics have been studied numerically by using phase diagrams, Lyapunov exponent spectrum, bifurcation diagram and extreme multistability can be found. In particular, the system can generate a complex bursting behavior in the presence of an external current. In order to verify the accuracy of the simulation, the phase diagram of FO heterogeneous neuron network is implemented by STM32 microcontroller, and results of the experiments are in great agreement with results of the numerical simulations. Finally, an image encryption and hiding method based on FO heterogeneous neuron network and discrete wavelet transform (DWT) is proposed. The experimental results demonstrate that the encryption and hiding scheme has excellent security and strong robustness.