A chaotic optical communication scheme assisted by artificial intelligence (AI) method is proposed. Different from other traditional solutions, the chaos synchronization of the proposed scheme is accomplished via an AI-based optoelectronic oscillator (AI-OEO) model and a QPSK driving signal distorted by chromatic dispersion. Specifically, the distorted driving signal excites the chaotic AI-OEO model to generate the chaotic signal for encryption and decryption. One significant advantage is that the strict dependence of chaos synchronization on physical devices can be reduced by using deep learning technology to model the actual experimental chaotic system. The proposed scheme has been experimentally validated in a chaotic-encrypted 256Gbit/s (32-GBaud) polarization-multiplexed 16QAM system over 1600km standard single-mode fiber (SSMF). The results show that the BER performance after 1600km transmission is lower than the 20% FEC threshold limit (BER = 2.4×10 -2 ). Furthermore, we conducted a comprehensive security assessment of the system, comparing it with traditional chaotic encryption schemes and performing a detailed analysis of the key space and key sensitivity. The proposed scheme is compatible with the existing high-speed coherent optical fiber communication systems, and the realization of precise chaos synchronization only requires ensuring that the driving signal is decoded without errors at the receiver. We believe that it has the potential to become a candidate solution for bidirectional high-speed secure communication with low complexity and cost.
27 December 2022 in Journal of Lightwave Technology
In this paper, a novel deep learning architecture, Fourier neural operator (FNO), has been introduced to ap-proximate the nonlinear Schrodinger equation which characterizes fiber transmission impairments such as fiber attenuation, chromatic dispersion, nonlinear impairments, etc. The proposed scheme is firstly verified via numerical simulation in 28-GBaud 4QAM/16QAM/64QAM systems over 1200-km standard single mode fiber (SSMF) under different launch powers. The simulation results show that the mean square errors (MSEs) are less than 0.002 and the effective SNRs differences between the proposed FNO and SSFM are all within 1 dB at 1200 km SSMF with the launching power from −5 dBm to 5 dBm. The performance of the proposed FNO model is further evaluated by a conceptual experiment. The experimental results show that the Q-factor performance of the FNO-based channel model is close to that of 1200 km experimental SSMF transmission link with the launching power from −3 dBm to 5 dBm.
Recently developed deep learning-based image inpainting methods have suggested their potential applications in filling large missing regions displaying plausible content. Nevertheless, most existing studies either merely considered the external information of the missing regions or relied on the region context to yield semantically coherent patches while overlooking the semantic relevance and feature continuity exhibited by missing regions; these demerits are likely to cause a discontinuous contextual structure and blurry texture details. In this work, a novel inside-outside attention layer (IOA) was proposed, capable of exploiting unmasked image features as references as well as learning the affinity between hole features to predict more consistent semantic information. The adversarial loss attributed exclusively on the natural image level cannot adequately generate a sharp texture detail. To address this problem, a texture component discriminator was introduced via wavelet decomposition to enhance the specific performance. Several experiments were performed on the CelebA and Places2 datasets. As revealed from the results, in contrast to the existing research, the proposed method is capable of restoring images with complex structures and significantly enhancing plausible structure and visual quality.
Fiber nonlinearity is a significant constraint on the maximum achievable capacity in long-distance fiber optic transmission systems. The fiber nonlinearity compensation can be done in digital signal processing through the inversion of the Manakov model using the split-step Fourier method (SSFM) with appropriate inverted channel parameters, which are implemented via digital backpropagation (DBP) algorithm. Nevertheless, the practical deployment of this algorithm and its variants require a significant amount of latency and signal processing resources. Currently, the data-driven approximation of partial differential equations (PDEs) is rapidly emerging as a powerful paradigm, wherein the operator learning framework has been demonstrated to effectively approximate the solution space of PDEs. Here, we construct an inverse parameterized Manakov model based on the neural operator, utilizing it to compensate the fiber nonlinearity impairments at low computational complexity. The model proposed has been validated in a 12,075 km wavelength division multiplexing (WDM) system. The received signal waveform compensated by this model produces results that are similar to single-channel DBP (optimal step size), with an MSE of 0.0008. The analysis of the computational complexity shows that the proposed algorithm achieves a performance loss of 0.024dB compared to single-channel DBP under the optimal step size, but its highly parallel nature significantly reduces computational latency.
16 January 2025 in Journal of Lightwave Technology
In long-haul and high-capacity coherent optical communications systems, it is essential to reduce the complexity of the underlying algorithms in the practical implementation of digital signal processing techniques for compensation of transmission impairments. Here, we investigate a factorized Fourier operator (F-FNO) based neural network (NN) for compensation the linear and nonlinear signal distortion in fiber simultaneously. The combination of the factorized Fourier transforms and residual connection enables the F-FNO based NN to compensate fiber impairments with higher accuracy, lower computational complexity and faster convergence rate, compared with our previous Fourier operator (FNO) based NN. The time consumption is further decreased by sharing the Fourier-domain weight matrices. The feasibility of the proposed scheme is verified experimentally in 32GBaud polarization division multiplexed (PDM) uniformly distributed 16/32/64-ary quadrature amplitude modulation (PDM-16/32/64QAM) as well as probabilistically shaped 16/32/64QAM (PDM-PS-16/32/64QAM) coherent optical communication systems. The experimental results demonstrate that the Q-factor improvement is extremely close to the standard digital back propagation (DBP) with 16 steps per span (16StPS-DBP) even in the highly nonlinear region, while with much lower computational complexity and time consumption.
We experimentally demonstrate a total net-rate of 27.88 Tb/s for C-band wavelength-division multiplexing (WDM) transmission over an ultralong span-length of 150 km. It is the largest net capacity × span-length product of 4182 Tb/s·km for C-band, single-core, standard single-mode optical fiber transmission over a length of more than 3,000 km. A total of 99 channels, spaced at 50 GHz intervals, are employed for transmitting 32 GBaud probabilistically constellation-shaped (PCS) 64QAM signals with an information entropy of 5.5. High gain amplifiers can achieve wavelength-division multiplexing (WDM) transmission with a bandwidth of 6.25 THz, at a noise figure below 4.3 dB, without the assistance of distributed Raman amplification.