{"id":126,"date":"2026-06-02T22:14:44","date_gmt":"2026-06-02T21:14:44","guid":{"rendered":"https:\/\/blogs.kent.ac.uk\/ipc\/?p=126"},"modified":"2026-09-18T10:28:10","modified_gmt":"2026-09-18T09:28:10","slug":"new-journal-of-lightwave-technology-paper-published-on-photonic-reservoir-computing-for-fmcw-lidar","status":"publish","type":"post","link":"https:\/\/blogs.kent.ac.uk\/ipc\/2026\/06\/02\/new-journal-of-lightwave-technology-paper-published-on-photonic-reservoir-computing-for-fmcw-lidar\/","title":{"rendered":"New Journal of Lightwave Technology Paper on All-Optical Input Masking for Photonic Reservoir Computing"},"content":{"rendered":"

We are delighted to announce that Dr Yuanli Yue<\/a>, a recent PhD graduate from the Photonics Lab at the University of Kent, has published a new research paper in the prestigious Journal of Lightwave Technology (JLT)<\/strong>, together with Prof. Hao Zhang from Nankai University and Dr Chao Wang<\/a> from Kent.<\/p>\n

The paper presents a photonic time-stretch reservoir computing system with input masking performed directly in the optical domain. Input masks are essential for enriching reservoir dynamics, but conventional electronically generated masks require high-bandwidth hardware and can limit processing speed.<\/p>\n

The researchers from our team experimentally demonstrated two optical masking schemes: a binary spectral mask using parallel diffraction gratings and a continuous chirp-coded mask generated by an unbalanced Mach-Zehnder interferometer. Photonic time stretch maps different wavelengths into time-domain virtual nodes, while nonlinear spectral mixing in a semiconductor optical amplifier couples the nodes and strengthens the reservoir dynamics.<\/p>\n

\u00a0\"\"<\/p>\n

The system was validated through waveform classification and spoken-digit recognition using the standard NIST TI-46 dataset. The diffraction-grating mask achieved 98.16% classification accuracy<\/strong>, while the Mach-Zehnder interferometer mask achieved 97.19%<\/strong>, compared with 80.65% without masking<\/strong>. The two designs offer complementary benefits: the diffraction-grating mask provides slightly better classification accuracy, whereas the continuous Mach-Zehnder mask produces a lower normalized mean-square error and improved temporal mapping.<\/p>\n

The work formed part of Dr Yue\u2019s PhD research at the University of Kent under the supervision of Dr Chao Wang, in collaboration with Dr Hao Zhang at Nankai University. It demonstrates how optical input masking can improve photonic reservoir computing while reducing reliance on high-speed electronic hardware, supporting future applications in ultrafast imaging, sensing and intelligent signal processing.<\/p>\n

This publication further strengthens the Photonics Lab\u2019s research activities in:<\/p>\n