| Memristors exhibit non-volatility and memory properties that can simulate biological neural synapses.Synaptic arrays composed of Memristors have a high degree of integration,and memristor neurons are becoming important building blocks for the implementation of neural-morphic computing and machine learning hardware.Moreover,the flexible electromagnetic constraints of Memristors can lead to complex phenomena in memristor synaptic neurons,such as chaotic discharges,which provide new implementation paths for chaos dynamics and neuron dynamics control.Due to the nonlinearity of Memristors,memristor neurons are prone to generating chaotic oscillations,and the resulting chaotic sequences are unpredictable,making them suitable for image encryption.Research has shown that the increased waveform complexity of memristor neuron chaotic systems enhances image encryption security.Thus,designing and implementing memristor chaotic neuron systems have both theoretical and engineering value.With the in-depth study of the control of complex dynamics in memristor systems,the regulation of memristor neuron attractors has become possible.In this article,we investigate the dynamic control of memristor Hindmarsh-Rose(HR)neurons.Firstly,we propose offset boosting for HR neurons that uses trigonometric and absolute value functions to achieve the coexistence management of multiple attractors.We introduce a offset constant term into the HR neuron model and hide the offset constant by introducing a periodic function,obtaining coexisting attractors of neuron morphology,and selecting different coexisting attractors in phase space by choosing different initial values.Secondly,we propose a rotation control method for memristor HR neurons that uses rotation matrices to adjust the amplitude and offset of attractors,enabling the rotation of attractors in phase space.We construct memristor neurons by introducing a locally active memristor in the first dimension of the HR neuron system,and then introduce a threedimensional matrix rotating around the z-axis.We use the rotation matrix to control the cluster firing and spike discharge of memristor neurons’ membrane potential and analyze the variation of the equilibrium point and characteristic values of the system and the variation law of the peak membrane potential with the angle change.Finally,we study the influence of the same memristor on the dynamic behavior of chaotic systems and neuron systems based on the electromagnetic constraints of Memristors.We use the chaotic output of the memristor system to enhance the anti-attack capability of color image encryption.We construct a three-dimensional memristor system with an embedded magneticcontrolled memristor in a controllable offset chaotic system,and a three-dimensional memristor neuron system with the same memristor introduced in the two-dimensional HR neuron,and study the effect of memristor computation on system dynamic behavior.We apply the corresponding chaotic output to image encryption and analyze the security of image encryption. |
