Retina-inspired photopic adaptive synaptic transistor with enhanced retention for in-sensor denoising and accurate recognition

With the growing demand for neuromorphic devices in intelligent perception systems, traditional artificial synaptic devices are facing challenges, as ambient light variations and background noise interference can severely degrade perception accuracy and deteriorate visual feature information. Here, inspired by the ultraviolet (UV) light perception of the bee retina, a novel organic optoelectronic synaptic transistor (OST) array capable of background noise suppression and photopic adaptability was developed. By adding an appropriate organic ionic salt, the device successfully exhibited a series of synaptic characteristics and information processing function. Further, a thin blocking layer was ingeniously designed, which improved the charge retention capability and non-volatility of the modified OST due to the dynamical modulation of plasticity. Comprehensive synapse characterizations were performed to validate the structure and plasticity, and high recognition accuracy of 91.64% was achieved by neuromorphic simulations. More importantly, an OST array was constructed and integrated into a convolutional neural network (CNN) to imitate photopic adaptation behavior of the bee retina in the UV region with noise interference. Despite a strong Gaussian noise (32%), an exceptional image recognition accuracy of 98.99% was obtained by using the OST array. The results indicate that the device provides excellent noise suppression performance and photopic adaptability for real-time intelligent vision perception in complex environments.