Review
Optics
Guoqing Wang, Yuan Zhou, Rui Min, E. Du, Chao Wang
Summary: Inspiring development in optical imaging has led to great applications in the science and engineering industry, particularly in the field of medical imaging. Photonic time-stretch imaging is an emerging innovation that has attracted widespread attention due to its ability to map space, wavelength, and time using dispersive medium in both spatial and time domains. The ultrafast imaging speed of this technique surpasses traditional methods by several orders of magnitudes, achieving an ultrahigh frame rate of tens of millions of frames per second. Moreover, it can combine various optical technologies like compressive sensing, nonlinear processing, and deep learning for ultrafast optical signal processing. This paper reviews the principle and recent developments of photonic time-stretch imaging and discusses future trends.
Article
Biochemical Research Methods
Siyuan Lin, Rubing Li, Yueyun Weng, Liye Mei, Chao Wei, Congkuan Song, Shubin Wei, Yifan Yao, Xiaolan Ruan, Fuling Zhou, Qing Geng, Du Wang, Cheng Lei
Summary: Imaging flow cytometry based on optical time-stretch imaging combined with a microfluidic chip is attracting attention in large-scale single-cell analysis due to its high throughput, precision, and label-free operation. Compressive sensing has been integrated to relieve the pressure of massive data, but image decompression adds computational overhead without generating additional information. In this work, a machine-learning network was proposed to analyze cells in the compressed domain without decompressing the images, achieving high-quality imaging and accurate cell classification at a compression ratio of 10%. This work provides a practical solution to the big data problem in optical time-stretch imaging flow cytometry.
JOURNAL OF BIOPHOTONICS
(2023)
Article
Optics
Yilin He, Yunhua Yao, Yu He, Zhengqi Huang, Pengpeng Ding, Dalong Qi, Zhiyong Wang, Tianqing Jia, Zhenrong Sun, Shian Zhang
Summary: Compressive wide-field fluorescence microscopy (CWFM) breaks the imaging speed limitation of conventional WFM by compressing images and using an iterative algorithm for reconstruction. It has important applications in microflow and cell tracking.
OPTICS AND LASERS IN ENGINEERING
(2023)
Article
Pharmacology & Pharmacy
Nejc Koracin, Matevz Zupancic, Franc Vrecer, Grega Hudovornik, Iztok Golobic
Summary: Liquid atomization plays a crucial role in pharmaceutical manufacturing, particularly in the production steps involving fluid bed granulation, tablet film coating, and pellet coating. This study focuses on characterizing the atomization process using high-speed imaging, optical microscopy, and continuous back-light illumination combined with robust image processing algorithms. The results provide insights into droplet size distribution and speed, as well as the correlation between droplet size and other process parameters. The proposed method demonstrates potential for real-time coating monitoring and rapid optimization of the atomization process.
INTERNATIONAL JOURNAL OF PHARMACEUTICS
(2022)
Article
Optics
Yirui Wang, Fengyi Jiang, Guohao Ju, Boqian Xu, Qichang An, Chunyue Zhang, Shuaihui Wang, Shuyan Xu
Summary: Segmented primary mirrors offer crucial advantages for constructing extra-large space telescopes, but the imaging quality is sensitive to phasing errors. Deep learning has been widely used for phasing segmented mirrors, with the advantages of high efficiency and avoiding stagnation issues. The proposed deep Bi-GRU neural network is effective for fine phasing of segmented mirrors, addressing the gradient vanishing problem and incorporating various errors for accurate and effective results.
Article
Computer Science, Information Systems
Zhufeng Suo, Youheng Dong, Fenghua Tong, Donghua Jiang, Xi Fang, Xiaoming Chen
Summary: In this paper, a novel image encryption scheme based on SSL-PUF and 2DCS is proposed. The scheme utilizes SSL-PUF for key distribution and 2DCS for compression and encryption, and optimizes the reconstruction method for high-quality image reconstruction. The proposed scheme demonstrates good performance.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Hardware & Architecture
Penta Anil Kumar, R. Gunasundari, R. Aarthi
Summary: In this paper, a novel deep-learning method is proposed for image reconstruction using a set of MRI, incorporating compressive sensing techniques to reduce image size and reconstruction time. By using convolution layers to simulate the process of compressed sampling and adopting Deep Convolutional Neural Network for image reconstruction, this method achieves promising results in experiments.
Article
Engineering, Electrical & Electronic
Lihao Zhuang, Liquan Shen, Zhengyong Wang, Yinyi Li
Summary: This paper proposes a novel priors guided adaptive underwater compressive sensing framework, dubbed UCSNet, which can effectively sample and reconstruct underwater images under a fixed low sampling ratio. The framework consists of three sub-networks: underwater priors extraction and guidance network, sampling matrix generation network, and channel-wise reconstruction network. Experimental results demonstrate that our framework outperforms other state-of-the-art methods in terms of underwater image reconstruction quality.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Engineering, Electrical & Electronic
Ljubisa Stankovic, Milos Brajovic, Danilo Mandic, Isidora Stankovic, Milos Dakovic
Summary: An improved coherence-based uniqueness relation for matching pursuit algorithms is proposed, along with a less conservative coherence index-based lower bound for signal sparsity. These results are generalized to the uniqueness condition of l(0)-norm minimization for signals represented in two orthonormal bases.
IEEE SIGNAL PROCESSING LETTERS
(2021)
Article
Computer Science, Information Systems
Xiuli Chai, Haiyang Wu, Zhihua Gan, Daojun Han, Yushu Zhang, Yiran Chen
Summary: A novel double color image encryption algorithm combining 2D compressive sensing with an embedding technique is proposed in this paper, achieving compression and encryption simultaneously and obtaining visually meaningful cipher images. Confusion of the compressed cipher images is done by index sequences generated from a 6D hyperchaotic system, while the relationship of the algorithm with plain images is enhanced by embedding feature parameters into the carrier image.
INFORMATION SCIENCES
(2021)
Article
Geochemistry & Geophysics
Han Gao, Guifeng Zhang, Min Huang
Summary: Compressive sensing (CS) is a super-resolution technique intensively studied in image acquisition and reconstruction. It has been applied to three-dimensional (3-D) laser imaging to enhance spatial resolution. However, existing approaches have limitations in terms of range resolution and long-range detection abilities. In this study, we propose a sampling model that combines CS sampling process with time-of-flight (TOF)-based range measurement procedure in pulsed-laser imaging. By exploring range information and introducing rank minimization and a fallback mechanism, our approach significantly improves the accuracy and quality of reconstructed 3-D images compared to recent methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Sidi Lu, Xin Yuan, Aggelos K. Katsaggelos, Weisong Shi
Summary: In this work, reinforcement learning is applied to video compressive sensing to adapt the compression ratio. The gap in previous studies of how to adapt B in the video SCI system is filled using RL. An RL model and various convolutional neural networks are employed to achieve adaptive sensing of video SCI systems. Additionally, the performance of an object detection network is utilized for RL-based adaptive video compressive sensing. This proposed adaptive SCI method can be implemented in low cost and real time, and takes the technology one step further towards real applications of video SCI.
ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY
(2023)
Article
Engineering, Mechanical
Hua-Ping Wan, Guan-Sen Dong, Yaozhi Luo
Summary: This paper proposes a new method to construct a dedicated dictionary for wind speed signals using the time-shift strategy, improving the performance of compressive sensing (CS) methodology in wind monitoring. The results demonstrate that the improved CS methodology outperforms the traditional CS algorithm and explores the influences of critical parameters comprehensively.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Geosciences, Multidisciplinary
Mengli Zhang, Yaoguo Li
Summary: A major challenge in magnetotelluric (MT) surveys is the limited number of stations due to cost and time constraints. We developed an efficient acquisition approach that uses sparse irregular patterns to optimize station layout, resulting in significant savings in cost and operational time while maximizing information from a given number of stations. Comparable information to a full regular grid can be obtained by using as few as 25% stations.
JOURNAL OF APPLIED GEOPHYSICS
(2023)
Review
Chemistry, Multidisciplinary
Mali Zhao, Dohyun Kim, Young Hee Lee, Heejun Yang, Suyeon Cho
Summary: This paper introduces critical advances in the field of quantum sensing of thermopower, ranging from atomic to several-hundred-nanometer scales, and discusses the roles of low-dimensionality, defects, spins, and relativistic effects in optimized power generation. Investigating the microscopic nature of thermopower in quantum materials can provide insights for the design of advanced materials for future thermoelectric applications, while quantum sensing techniques for thermopower can pave the way for practical and novel energy devices towards a sustainable society.
ADVANCED MATERIALS
(2023)
Editorial Material
Engineering, Electrical & Electronic
Roberto Sabella, David Plant, Hongwei Chen, Antonella Bogoni, Vladimir Stojanovic
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)
Article
Engineering, Electrical & Electronic
Yubin Zang, Zhenming Yu, Kun Xu, Xingzeng Lan, Minghua Chen, Sigang Yang, Hongwei Chen
Summary: This paper proposes a principle-driven fiber transmission model based on physical induced neural network (PINN) which views fiber transmission as an equation solving problem. By considering the physical principles and initial conditions, this model can effectively handle the prediction tasks of pulse evolution, signal transmission, and fiber birefringence for different transmission parameters of fiber telecommunications.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)
Article
Engineering, Electrical & Electronic
Shuai Shao, Jiachen Li, Hongwei Chen, Sigang Yang, Minghua Chen
Summary: This study presents a hybrid integrated gain-switched optical frequency comb source with eight pure continuous comb lines, a narrow linewidth, and a large frequency spacing adjustment range. By coupling a silicon nitride microring reflector to a commercially available distributed feedback laser, the self-injection locking effect greatly enhances the carrier-to-noise ratio and phase correlation between comb lines. This cost-efficient and robust comb source has potential applications in areas such as radio over fiber and coherent optical communication.
IEEE PHOTONICS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Liwei Tang, Sigang Yang, Hongwei Chen, Minghua Chen
Summary: In this study, a heterodyne hybrid optical phase-locked loop system was proposed to reduce the spectral linewidth of the slave laser and shorten the loop delay. By packaging a self-injection locked semiconductor laser, photodiode, and electronic circuit, the slave laser achieved a narrow spectral linewidth of 2.8 kHz. The utilization of a DFB laser diode coupled to a microring resonator on a Si3N4 integrated waveguide platform allowed for a shortened optical path delay to 30 ps. This resulted in a reduction of residual loop noise and phase noise variance.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)
Article
Optics
Wanxin Shi, Zheng Huang, Honghao Huang, Chengyang Hu, Minghua Chen, Sigang Yang, Hongwei Chen
Summary: This study proposes a novel lensless opto-electronic neural network architecture for machine vision applications. By optimizing a passive optical mask through task-oriented neural network design, the device size and calculation required are reduced. Experimental results demonstrate good performance in handwritten digit classification and face recognition tasks.
LIGHT-SCIENCE & APPLICATIONS
(2022)
Article
Optics
Jiachen Li, Sigang Yang, Hongwei Chen, Xingjun Wang, Minghua Chen, Weiwen Zou
Summary: In this paper, a fully integrated hybrid microwave photonic receiver (FIH-MWPR) is presented, which combines indium phosphide (InP) laser chip and monolithic silicon-on-insulator (SOI) photonic circuit into the same substrate. The integrated receiver exhibits compact size and low power consumption, and supports a wide operation bandwidth. It has been successfully used for processing radar signals in different frequency bands and demonstrated high-resolution-ranging capability.
PHOTONICS RESEARCH
(2022)
Article
Optics
Honghao Huang, Chengyang Hu, Jingwei Li, Xiaowen Dong, Hongwei Chen
Summary: This paper proposes a novel deep learning framework CoCoCs for compressive imaging, which optimizes the recovery algorithm with optical coding to improve the quality of reconstruction. Experimental results demonstrate that CoCoCs can generate realistic images and videos suitable for both human and computer vision.
Article
Optics
H. O. N. G. H. A. O. Huang, J. I. A. J. I. E. Teng, Y. U. Liang, C. H. E. N. G. Y. A. N. G. Hu, M. I. N. G. H. U. A. Chen, S. I. G. A. N. G. Yang, H. O. N. G. W. E. Chen
Summary: This study proposes a new method for compressive video sensing, called KH-CVS, which utilizes a key frames assisted hybrid encoding paradigm to improve video reconstruction quality by capturing short-exposure key frames without coding and long-exposure encoded compressive frames.
Article
Optics
Yubin Zang, Zhenming Yu, Kun Xu, Minghua Chen, Sigang Yang, Hongwei Chen
Summary: In this paper, a data-driven fiber model based on a deep neural network with multi-head attention mechanism is proposed. This model, which predicts signal evolution in optical fiber telecommunications, offers advantages in computation time without sacrificing accuracy compared to the conventional split-step fourier method. Unlike other neural network models, this model achieves a relatively good balance between prediction accuracy and distance generalization, particularly when higher bit rates and more complex modulation formats are utilized. Numerical demonstrations demonstrate this model's ability to predict 16-QAM 160Gbps signals with transmission distances ranging from 0 to 100 km, with or without noise.
Article
Multidisciplinary Sciences
Tingzhao Fu, Yubin Zang, Yuyao Huang, Zhenmin Du, Honghao Huang, Chengyang Hu, Minghua Chen, Sigang Yang, Hongwei Chen
Summary: Researchers have developed an on-chip diffractive optical neural network (DONN) using 1D dielectric metasurfaces, achieving 90% classification accuracy, computing at 10^16 flops/mm^2, and consuming 10E-17 J/Flop. The proposed DONN is based on a silicon-on-insulator platform and offers high integration and low power consumption for machine learning tasks.
NATURE COMMUNICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Xingchen He, Lianshan Yan, Lin Jiang, Anlin Yi, Zhengyu Pu, Youren Yu, Hongwei Chen, Wei Pan, Bin Luo
Summary: This paper introduces a novel deep learning architecture, Fourier neural operator (FNO), to approximate the nonlinear Schrodinger equation for characterizing various fiber transmission impairments. The proposed scheme is validated through numerical simulation in 28-GBaud systems over 1200-km standard single mode fiber (SSMF) with different launch powers. The simulation results demonstrate that the proposed FNO achieves low mean square errors (MSEs) and maintains effective signal-to-noise ratios (SNRs) comparable to SSFM within 1 dB difference.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2023)
Article
Optics
Wencan Liu, Tingzhao Fu, Yuyao Huang, Run Sun, Sigang Yang, Hongwei Chen
Summary: A new method is proposed to improve the integration level of an on-chip diffractive optical neural network (DONN) based on a standard silicon-on-insulator (SOI) platform. By using subwavelength silica slots as the hidden layer, a metaline with high computation capacity is formed. The deep mapping regression model (DMRM) accurately characterizes the propagation process of light in the metalines, enhancing the integration level and eliminating the need for approximate conditions. The method is validated on the Iris plants dataset, achieving a testing accuracy of 93.3%. This research provides a potential solution for future large-scale on-chip integration.
Article
Engineering, Electrical & Electronic
Ziyang Lu, Jiachen Li, Yunhao Wu, Hongwei Chen, Sigang Yang, Minghua Chen
Summary: To meet the requirements of processing radio frequency signals across different frequency bands, a reconfigurable microwave photonic bandpass filter is demonstrated by modulating the phase and flipping the optical spectrum of a cascaded microring resonator (MRR) filter pool chip. The performance of the established microwave photonic filter (MPF) is evaluated using the characterized cascaded MRR filter pool chip on the Si3N4 platform. The center frequency of the demonstrated MPF system can be tuned within 5.8-18.2 GHz, with a bandwidth of 2.1-3.5 GHz and a response shape factor of approximately 2.2. With its reconfigurability and good shape factor, the demonstrated reconfigurable RF bandpass filter has potential applications in future software-defined RF frontends.
IEEE PHOTONICS JOURNAL
(2023)
Article
Optics
Yuyao Huang, Tingzhao Fu, Honghao Huang, Sigang Yang, Hongwei Chen
Summary: Photonic integrated circuits enable efficient optical computing with the optical convolution unit (OCU), achieving high accuracy and clarity in deep learning tasks. OCU offers low energy consumption and high information density, providing a parallel lightweight solution for handling high-dimensional tensors in the future.
PHOTONICS RESEARCH
(2023)
Article
Engineering, Electrical & Electronic
Yubin Zang, Zhenming Yu, Kun Xu, Minghua Chen, Sigang Yang, Hongwei Chen
Summary: This paper presents a novel fiber transmission model using cascaded neural networks and multi-head attention mechanism to solve signal transmission prediction problems in multi-span long-haul fiber link. The model can accurately predict the signal transmission results with low time cost, compared to traditional methods.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)