Abstract: A special complex hysteresis characteristic exists in Voice Coil Motor (VCM) under high-frequency, high-speed and high-acceleration, which makes the system performance go worse. In order to achieve the precise positioning control for VCM, a hybrid hysteresis model which consists of the improved Preisach units and dynamic neural network was proposed. In the improved Preisach model, the non-monotonic information was introduced for describing the non-monotonic hysteresis behavior. Based on the former part, a dynamic radial basis function (RBF) neural network was applied to compensate the amplitude and phase error between the dynamic Preisach output and the real VCM hysteresis for a higher accuracy. Finally, the simulation experimental results based on the measured data show that the proposed hybrid model is effective not only for the general hysteresis modeling and but also for the non-monotonic complex hysteresis. Comparing with the dynamic neural network modeling, this hybrid model is of a higher modeling accuracy.