Self-powered image reading and computing method based on integration of photodiode and memristor

Abstract

This invention belongs to the field of biomimetic visual perception technology and relates to a self-powered image reading and calculation method based on the integration of photodiodes and memristors. This method directly applies the photovoltage generated by the photodiode to the memristor, utilizing the generated self-powered current to achieve image reading and calculation. During image reading, the light intensity of the external light source is not less than the self-powered reading threshold. The self-powered current output from each row and unit in the array is collected, and the resistance value of each unit's memristor is calculated to reconstruct the stored image information. During image calculation, the weights calculated by the neural network in the memristor are stored in the memristor resistance value. The light intensity of the external light source is not less than the sum of the self-powered reading threshold and the additional light intensity. The self-powered currents output from different columns are accumulated to obtain the calculated image. This invention's method requires no external electrical signal input; the input optical signal simultaneously achieves both power supply and information input, which can greatly reduce the power consumption of information processing.

Images