Programmable ferroelectric bionic vision hardware with selective attention for high-precision image classification

Selective attention is an efficient processing strategy to allocate computational resources for pivotal optical information. However, hardware implementation of selective visual attention in conventional intelligent systems is usually bulky and complex with high computational cost. Here, programmable ferroelectric bionic vision hardware to emulate selective attention is proposed. The tunneling effect of photogenerated carriers is controlled by dynamic variation of energy barrier, enabling modulation of memory strength from 9.1% to 47.1% without peripheral storage. Molecular polarization of ferroelectric P(VDF-TrFE) provides multiple nonvolatile states and selective attention in a single device. Arrayed devices selectively record UV information and suppress background with high current decibel level. The positively polarized device shows high wavelength dependence in image attention processing, and the ferroelectric sensory network attains 95.7% accuracy in pattern classification of multi-wavelength images. This enriches neuromorphic functions of bioinspired sensing devices and paves the way for future bioinspired optoelectronics.

Rengjian Yu, Lihua He, Changsong Gao, Xianghong Zhang, Enlong Li, Tailiang Guo, Wenwu Li, Huipeng Chen