Fiber neural networks for the intelligent optical fiber communication signal processing

With the rapid growth of global data traffic, enhancing the intelligence and efficiency of optical fiber communication systems has become even more crucial. Current intelligent signal processing requires converting optical signals to electrical ones, leading to high latency and power consumption. To that end, a research collaboration led by Tsinghua University proposed a novel solution: a fiber neural network that processes information entirely within the optical domain.

“The system architecture consists of an input branch, an output branch, and an optical computing loop,” explains corresponding author Hongwei Chen. “Key components include a laser source, single-mode fiber, erbium-doped fiber amplifier, modulators, and dispersion-compensating fiber.”

The optical computing loop physically implements the linear operations of a neural network by using time-stretched optical pulses to perform vector-matrix multiplication—the core calculation between neural network layers.

“We applied the system to modulation format recognition, a fundamental task in optical communications. The fiber neural network successfully identified three modulation formats: OOK, PAM, and PSK,” shares Chen.

Notably, under ideal noise-free conditions, the system achieved 100% classification accuracy. It also maintained strong robustness when experimental noise from amplifiers and detectors was introduced, demonstrating its potential for real-world deployment.

The team’s findings, published in iOptics, show that this fiber neural network framework can execute AI-driven computations directly using light.

“Importantly, this method avoids frequent optical-electrical conversions, thereby reducing processing delay and energy use,” adds Chen.

Beyond modulation recognition, the approach may be applicable to other intelligent optical communication applications such as fault detection and channel modeling, paving the way for deeper integration of artificial intelligence and photonic technology.

###

Contact the author: 

Hongwei Chen, Department of Electronic Engineering, Tsinghua University, Beijing, China
Email: chenhw@tsinghua.edu.cn

The publisher KeAi was established by Elsevier and China Science Publishing & Media Ltd to unfold quality research globally. In 2013, our focus shifted to open access publishing. We now proudly publish more than 200 world-class, open access, English language journals, spanning all scientific disciplines. Many of these are titles we publish in partnership with prestigious societies and academic institutions, such as the National Natural Science Foundation of China (NSFC).