High-speed optical target feature extractor based on multimode fiber optical neural network

Authors: Hu, Liuyang; Qiao, Yingtai; Lin, Wei; Zhang, Hao; Liu, Bo

Source journal: Optics express (2026-11-16)

DOI: 10.1364/OE.613930

Abstract

With the widespread application of artificial intelligence, the requirement for the signal processing speed has been dramatically increasing. Optical neural networks, benefiting from intrinsic parallelism and high bandwidth, offer a promising pathway for high-speed information processing. Here, we proposed a multimode fiber (MMF)-based optical neural network (ONN) for high-speed target feature extraction, which utilizes the MMF as a high-dimensional optical encoder. Through complex mode coupling, the input image of the target is transformed into structured optical feature representations, enabling ultrafast and inherently parallel feature extraction. A 16-core fan-out device is employed to spatially sample the resulting speckle field, producing a compact 4 x 4 feature matrix for efficient downstream processing. The extracted features are evaluated using both image reconstruction and classification tasks, which respectively quantify information fidelity and discriminative capability. Furthermore, high-speed dynamic modulation experiments are conducted to investigate the stability and invariance of the encoded features under realistic operating conditions. Experimental results demonstrate that the proposed framework achieves high reconstruction accuracy and robust classification performance, while maintaining stable feature representations under dynamic perturbations. These findings establish MMFs as effective and scalable optical front-end processors, offering a promising route toward real-time optical neural computing.

Read the full text

Related topics

Related open access articles

Abstract and bibliographic data from PubMed (U.S. National Library of Medicine). Copyright of the abstract belongs to the original publisher or authors.