Time-stretch enabled multifunctional incoherent optical processor for analog-digital computing

Optical information processing systems have attracted considerable attention in various signal sensing applications. The development of multifunctional architectures is crucial for extending their operational versatility. This study presents a multifunctional incoherent optical information processing system underpinned by time-stretch technology, engineered to execute convolution operations for optical neural networks while simultaneously demonstrating its applicability in microwave photonics through pulse compression tasks. Experimental validation shows the system enables a convolutional neural network implementation achieving 97.0% recognition accuracy on eight modulation formats from the RML2016.10a dataset in high-SNR scenarios. Notably, it also demonstrates practical feasibility for pulse compression with a 10.6-fold compression factor attained for linear frequency-modulated signals. These advancements provide new perspectives on the multifaceted applications of optical processing paradigms, particularly highlighting the potential of the system in next-generation intelligent communication architectures requiring joint analog-digital processing.