Artificial Shape Perception Retina Network Based on Tunable Memristive Neurons

Lin Bao,Yuchao Yang,Zongwei Wang,Yimao Cai,Jian Kang,Ru Huang,Zhizhen Yu,Yichen Fang

doi:10.1038/s41598-018-31958-6

Lin Bao, Yuchao Yang + Show 6 more

Open Access

https://doi.org/10.1038/s41598-018-31958-6

Copy DOI

Abstract

Retina shows an extremely high signal processing efficiency because of its specific signal processing strategy which called computing in sensor. In retina, photoreceptor cells encode light signals into spikes and ganglion cells finish the shape perception process. In order to realize the neuromorphic vision sensor, the one-transistor-one-memristor (1T1M) structure which formed by one memristor and one MOSFET in serial is used to construct photoreceptor cell and ganglion cell. The voltage changes between two terminals of memristor and MOSFET can mimic the changes of membrane potential caused by spikes and illumination respectively. In this paper, the tunable memristive neurons with 1T1M structures are built. According to the concept of receptive field of ganglion cells (GCs) in the retina, the artificial shape perception retina network is constructed with these memristive neurons. The final results show that the artificial retina can extract shape information from the image and transfer it into spike frequency realizing the function of computing in sensor.

Highlights

There are two types of photoreceptor cells in the retina: cone cells and rod cells[3,4]
The simulation results show that the artificial shape perception retina network can extract shape information from the image and transform it into spike frequency realizing the function of computing in sensor
We proposed a new type of memristive neuron based on one-transistor-one-memristor (1T1M) structure which can integrate spike signal and analog signal at same time

Summary

Introduction

There are two types of photoreceptor cells in the retina: cone cells and rod cells[3,4]. Because of illumination will decrease the concentration of cGMP, it is harder for photoreceptor cells to create spikes when they are exposed in light. On the cell level, a tunable artificial neuron needs to be constructed in order to mimic photoreceptor cells and ganglion cells. On the network level, in order to realize the receptive field of ganglion cell, the connections between photoreceptors and ganglion cell need to be constructed. A novel tunable integrate-and-fire memristive neuron was built on 1T1M structure With these neurons, we mimicked the behaviour of photoreceptor cells and demonstrated the concept of receptive field of ganglion cells (GCs). The simulation results show that the artificial shape perception retina network can extract shape information from the image and transform it into spike frequency realizing the function of computing in sensor

Methods

Discussion

Conclusion