Time-domain low-complexity clock recovery for non-integer oversampled Nyquist signals with a small roll-off factor.

A clock recovery algorithm (CRA) suitable for non-integer oversampled Nyquist signals with a small roll-off factor (ROF) is appealing to short-reach high-speed inter-datacenter transmission systems which need to cut down the transceiver power consumption and cost by reducing the oversampling factor (OSF) and using cheap low-bandwidth components. However, due to the lack of a suitable timing phase error detector (TPED), CRAs proposed now fail for non-integer OSFs below two and small ROFs close to zero and are not hardware-efficient. To solve these problems, we propose a low-complexity TPED by modifying the time-domain quadratic signal and reselecting the synchronization spectral component. We demonstrate that the proposed TPED, in combination with a piece-wise parabolic (PWP) interpolator, can significantly improve the performance of feedback CRAs for non-integer oversampled Nyquist signals with a small ROF. Numerical simulations and experiments show that, based on the improved CRA, the receiver sensitivity penalty can keep below 0.5 dB when the OSF is reduced from 2 to 1.25 and the ROF is varied from 0.1 to 0.001 for 45 GBaud dual-polarization Nyquist 16QAM signals.

High receiver skew-tolerant and hardware-efficient clock recovery for short-reach coherent transmission.

The performance of the high-baud-rate and high-order-modulation-format short-reach coherent transmission systems is sensitive to the in-phase and quadrature (IQ) skew. The conventional receiver IQ skew compensation schemes based on adaptive equalizers (AEQs) suffer from the IQ skew enhanced timing jitter incurred by the clock recovery algorithm (CRA), resulting in a serious sensitivity degradation. In this paper, we first propose a novel multiplication-free timing phase error detector (TPED) with the gain insensitive to the receiver IQ skew and the capability to deal with the complex-valued Nyquist signal with an arbitrary roll-off factor and its real-valued IQ tributaries. Based on the TPED, we then propose a new all-digital feedback CRA able to compensate for the receiver IQ skew. With the novel CRA, the IQ skew enhanced timing jitter is eliminated and the receiver sensitivity can be improved by more than 1 dB for the 61 GBaud dual-polarization Nyquist 16QAM system for an IQ skew of 5 ps. Furthermore, the proposed CRA can reduce the computation complexity of the AEQ by more than 25% compared with the existing schemes by relieving the AEQ from IQ skew compensation. Both numerical simulations and experiments are carried out to validate the advantages of the proposed algorithms. The high-skew-tolerant and low-complexity CRA is a strong candidate for the power-sensitive high-speed short-reach coherent transmission systems.

Simultaneously precise frequency response and IQ skew calibration in a self-homodyne coherent optical transmission system.

The self-homodyne coherent detection (SHCD) system is becoming more popular in intra-data center applications nowadays. However, for a high-speed SHCD system, the device imperfection such as transmitter (Tx) and receiver (Rx) side in-phase (I)/quadrature-phase (Q) time skew and bandwidth limitation will greatly restrict the transmission performance. The current mainstream calibration methods for traditional optical transceivers rely on the effect of frequency offset and phase noise to separate the Tx and Rx imperfection, which is not compatible with the SHCD system. In this paper, we have proposed and demonstrated a highly precise calibration method that can be applied in dual-polarization (DP) SHCD system. Based on the specially designed multi-tone signals, the amplitude/phase frequency response (AFR/PFR) of the transceiver and the Tx/Rx IQ skew can be obtained by just one measurement even after long-distance fiber transmission. By using a 4 MHz linewidth distributed feedback (DFB) laser, a DP SHCD transmission system combined with a 20 GHz optical transceiver and two 10 km standard single-mode fibers is experimentally constructed. The test results indicate that the measurement error of the AFR/PFR and Tx/Rx skew are within ±1dB/±0.15rad and ±0.3ps respectively, and the dynamic range for IQ skew calibration can reach dozens of picoseconds. The measured bit error rate value of 46GBaud DP-16QAM signals/35GBaud DP-64QAM signals are improved from 2.30e-2 to 2.18e-3/9.59e-2 to 2.20e-2 with the help of the proposed calibration method.

iFlowGAN: An Invertible Flow-Based Generative Adversarial Network for Unsupervised Image-to-Image Translation.

We propose iFlowGAN that learns an invertible flow (a sequence of invertible mappings) via adversarial learning and exploit it to transform a source distribution into a target distribution for unsupervised image-to-image translation. Existing GAN-based generative model such as CycleGAN [1], StarGAN [2], AGGAN [3] and CyCADA [4] needs to learn a highly under-constraint forward mapping F: X → Y from a source domain X to a target domain Y. Researchers do this by assuming there is a backward mapping B: Y → X such that x and y are fixed points of the composite functions B °F and F °B. Inspired by zero-order reverse filtering [5], we (1) understand F via contraction mappings on a metric space; (2) provide a simple yet effective algorithm to present B via the parameters of F in light of Banach fixed point theorem; (3) provide a Lipschitz-regularized network which indicates a general approach to compose the inverse for arbitrary Lipschitz-regularized networks via Banach fixed point theorem. This network is useful for image-to-image translation tasks because it could save the memory for the weights of B. Although memory can also be saved by directly coupling the weights of the forward and backward mappings, the performance of the image-to-image translation network degrades significantly. This explains why current GAN-based generative models including CycleGAN must take different parameters to compose the forward and backward mappings instead of employing the same weights to build both mappings. Taking advantage of the Lipschitz-regularized network, we not only build iFlowGAN to solve the redundancy shortcoming of CycleGAN but also assemble the corresponding iFlowGAN versions of StarGAN, AGGAN and CyCADA without breaking their network architectures. Extensive experiments show that the iFlowGAN version could produce comparable results of the original implementation while saving half parameters.

Fast and simple calibration of frequency response and IQ skew for a coherent optical transmitter using a low-bandwidth photodetector.

A fast, precise, and low-cost coherent optical transmitter calibration scheme is proposed that uses multi-tone signals of a novel, to the best of our knowledge, design with unequal frequency intervals. With a single measurement, the proposed scheme can simultaneously calibrate the frequency response and the IQ skew of the transmitter using only a low-bandwidth photodiode. Simulation and experimental results indicate that the measurement error in the frequency response and coherent transmitter (Tx) skew is less than 0.3 dB and 0.2 ps, respectively. The feasibility of the proposed scheme is verified by an experiment involving 25 Gbaud 16-quadrature amplitude modulation (QAM) signal transmission using a Kramers-Kronig (KK) receiver. With the help of this calibration method, the measured bit error ratio performance was increased from 1.77e-2 to 3.52e-3 when the received optical power was -8 dBm.

DSP-free remote antenna unit in a coherent radio over fiber mobile fronthaul for 5G mm-wave mobile communication.

We propose a novel coherent analog radio over fiber (A-RoF) scheme to realize the generation, separation, and detection of four-independent mm-wave signals with the same carrier frequency on a single-wavelength for 5th generation (5G) mobile communication, and no digital signal processing (DSP) algorithms are required in remote antenna unit (RAU). In baseband unit (BBU), four-independent mm-wave signals are modulated on the two orthogonal polarization states of a single wavelength based on a dual-polarization IQ modulator using the dual single-sideband (SSB) modulation and polarization division multiplexing (PDM) technique. In RAU, a novel carrier polarization rotation module based on the self-polarization stabilization technique is proposed, and thus the four-independent mm-wave signals can be detected by self-coherent detection. Besides, the power fading effect induced by the chromatic dispersion could be overcome thanks to the optical SSB modulation, contributing to the increased coverage. By these means, no DSP algorithms are required in RAU, and the latency of signal processing could be significantly reduced. The experimental results show our proposed scheme could support 38.4 Gbps 16-ary quadrature amplitude modulation (16QAM) signals at 14 GHz over 30 km standard single-mode fiber (SSMF) transmission without DSP, satisfying 3rd Generation Partnership Project (3GPP) requirements. Besides, the measured error vector magnitude (EVM) value of 800 MBaud 16QAM signals at 28 GHz over 50 km SSMF transmission is 12.99%. This research provides a potential solution for the 5G mobile fronthaul.

Asymmetric dual-SSB modulation for photonic co-frequency mm-wave signals generation and DSP-free receiver.

An asymmetric dual-single-sideband (SSB) modulation scheme for photonic co-frequency millimeter (mm)-wave signals generation and digital signal processing (DSP)-free receiver is experimentally demonstrated for the first time, to the best of our knowledge. To effectively avoid the sideband crosstalk in the traditional symmetric dual-SSB modulation scheme, not only two vector-modulated signals but also two unmodulated sidebands are modulated on the two asymmetric sides of an optical carrier in this scheme. An optical delay line interferometer could easily separate these two asymmetric dual-SSB signals simultaneously in the receiver, and thus the photonic frequency up-conversion is realized. Besides, this scheme is free of dispersion-induced RF power fading thanks to the SSB modulation. By this means, no digital compensation algorithms such as carrier phase recovery, fiber dispersion compensation, and channel equalization are required, contributing to the DSP-free receiver. In our experiment, two 32 GHz 3.2 Gb/s 16-ary quadrature amplitude modulation mm-wave signals are produced using two RF signals with the carrier frequencies of 12 GHz and 20 GHz. The error vector magnitude (EVM) performances of these two mm-wave signals after 25.5 km standard single-mode fiber transmission are better than 3rd Generation Partnership Project requirements without using any digital compensation algorithms.

Robust Kernelized Multiview Self-Representation for Subspace Clustering.

In this article, we propose a multiview self-representation model for nonlinear subspaces clustering. By assuming that the heterogeneous features lie within the union of multiple linear subspaces, the recent multiview subspace learning methods aim to capture the complementary and consensus from multiple views to boost the performance. However, in real-world applications, data feature usually resides in multiple nonlinear subspaces, leading to undesirable results. To this end, we propose a kernelized version of tensor-based multiview subspace clustering, which is referred to as Kt-SVD-MSC, to jointly learn self-representation coefficients in mapped high-dimensional spaces and multiple views correlation in unified tensor space. In view-specific feature space, a kernel-induced mapping is introduced for each view to ensure the separability of self-representation coefficients. In unified tensor space, a new kind of tensor low-rank regularizer is employed on the rotated self-representation coefficient tensor to preserve the global consistency across different views. We also derive an algorithm to efficiently solve the optimization problem with all the subproblems having closed-form solutions. Furthermore, by incorporating the nonnegative and sparsity constraints, the proposed method can be easily extended to a useful variant, meaning that several useful variants can be easily constructed in a similar way. Extensive experiments of the proposed method are tested on eight challenging data sets, in which a significant (even a breakthrough) advance over state-of-the-art multiview clustering is achieved.

Hyper-Laplacian Regularized Multilinear Multiview Self-Representations for Clustering and Semisupervised Learning.

In this paper, we address the multiview nonlinear subspace representation problem. Traditional multiview subspace learning methods assume that the heterogeneous features of the data usually lie within the union of multiple linear subspaces. However, instead of linear subspaces, the data feature actually resides in multiple nonlinear subspaces in many real-world applications, resulting in unsatisfactory clustering performance. To overcome this, we propose a hyper-Laplacian regularized multilinear multiview self-representation model, which is referred to as HLR-M2VS, to jointly learn multiple views correlation and a local geometrical structure in a unified tensor space and view-specific self-representation feature spaces, respectively. In unified tensor space, a well-founded tensor low-rank regularization is adopted to impose on the self-representation coefficient tensor to ensure global consensus among different views. In view-specific feature space, hypergraph-induced hyper-Laplacian regularization is utilized to preserve the local geometrical structure embedded in a high-dimensional ambient space. An efficient algorithm is then derived to solve the optimization problem of the established model with theoretical convergence guarantee. Furthermore, the proposed model can be extended to semisupervised classification without introducing any additional parameters. An extensive experiment of our method is conducted on many challenging datasets, where a clear advance over state-of-the-art multiview clustering and multiview semisupervised classification approaches is achieved.

Interpreting and Extending the Guided Filter via Cyclic Coordinate Descent.

The guided filter (GF) is a widely used smoothing tool in computer vision and image processing. However, to the best of our knowledge, few papers investigate the mathematical connection between this filter and the least-squares optimization. In this paper, we first interpret the guided filter as the cyclic coordinate descent (CCD) solver of a least-squares objective function. This discovery implies an extension approach to generalize the guided filter since we can change the least-squares objective function and define new filters as the first pass iteration of the CCD solver of modified objective functions. In addition, referring to the iterative minimizing procedure of the CCD, we can derive new rolling filtering schemes. So, we are reasonable to say that our discovery not only reveals an approach to design new GF-like filters adapting to specific requirements of applications but also offers thorough explanations for two rolling filtering schemes of the guided filter as well as the method to extend them. Experiments prove our new proposed filters and rolling filtering schemes could produce state-of-the-art results.

Speeding Up the Bilateral Filter: A Joint Acceleration Way.

Computational complexity of the brute-force implementation of the bilateral filter (BF) depends on its filter kernel size. To achieve the constant-time BF whose complexity is irrelevant to the kernel size, many techniques have been proposed, such as 2D box filtering, dimension promotion, and shiftability property. Although each of the above techniques suffers from accuracy and efficiency problems, previous algorithm designers were used to take only one of them to assemble fast implementations due to the hardness of combining them together. Hence, no joint exploitation of these techniques has been proposed to construct a new cutting edge implementation that solves these problems. Jointly employing five techniques: kernel truncation, best N-term approximation as well as previous 2D box filtering, dimension promotion, and shiftability property, we propose a unified framework to transform BF with arbitrary spatial and range kernels into a set of 3D box filters that can be computed in linear time. To the best of our knowledge, our algorithm is the first method that can integrate all these acceleration techniques and, therefore, can draw upon one another's strong point to overcome deficiencies. The strength of our method has been corroborated by several carefully designed experiments. In particular, the filtering accuracy is significantly improved without sacrificing the efficiency at running time.