Searcharxiv⌕ Search

arXiv subjects

Ling Xu

Publications and source records attributed to Ling Xu.

33 records · Page 2Linked to original sources
arXiv2020

PPG-based singing voice conversion with adversarial representation learning

Singing voice conversion (SVC) aims to convert the voice of one singer to that of other singers while keeping the singing content and melody. On top of recent voice conversion works, we propose a novel model to steadily convert songs while keeping their naturalness and intonation. We build an end-to-end architecture, taking phonetic posteriorgrams (PPGs) as inputs and generating mel spectrograms. Specifically, we implement two separate encoders: one encodes PPGs as content, and the other compresses mel spectrograms to supply acoustic and musical information. To improve the performance on timbre and melody, an adversarial singer confusion module and a mel-regressive representation learning module are designed for the model. Objective and subjective experiments are conducted on our private Chinese singing corpus. Comparing with the baselines, our methods can significantly improve the conversion performance in terms of naturalness, melody, and voice similarity. Moreover, our PPG-based method is proved to be robust for noisy sources.

cs.SD↗
arXiv2020

Precoding and Transmit Antenna Subarray Selection for Secure Hybrid Spatial Modulation

Spatial modulation (SM) is a particularly important form of multiple-input-multiple-output (MIMO). Unlike traditional MIMO, it uses both modulation symbols and antenna indices to carry information. In this paper, to avoid the high cost and circuit complexity of fully-digital SM, we mainly consider the hybrid SM system with a hybrid precoding transmitter architecture, combining a digital precoder and an analog precoder. Here, the partially-connected structure is adopted with each radio frequency chain (RF) being connected to a transmit antenna subarray (TAS). In such a system, we made an investigation of secure hybrid precoding and transmit antenna subarray selection (TASS) methods. Two hybrid precoding methods, called maximizing the approximate secrecy rate (SR) via gradient ascent (Max-ASR-GA) and maximizing the approximate SR via alternating direction method of multipliers (Max-ASR-ADMM), are proposed to improve the SR performance. As for TASS, a high-performance method of maximizing the approximate SR (Max-ASR) TASS method is first presented. To reduce its high complexity, two low-complexity TASS methods, namely maximizing the eigenvalue (Max-EV) and maximizing the product of signal-to-interference-plus-noise ratio and artificial noise-to-signal-plus-noise ratio (Max-P-SINR-ANSNR), are proposed. Simulation results will demonstrate that the proposed Max-ASR-GA and Max-ASR-ADMM hybrid precoders harvest substantial SR performance gains over existing method. For TASS, the proposed three methods Max-ASR, Max-EV, and Max-P-SINR-ANSNR perform better than existing leakage method. Particularly, the proposed Max-EV and Max-P-SINR-ANSNR is low-complexity at the expense of a little performance loss compared with Max-ASR.

eess.SP↗
arXiv2019

Secure Hybrid Digital and Analog Precoder for mmWave Systems with low-resolution DACs and finite-quantized phase shifters

Millimeter wave (mmWave) communication has been regarded as one of the most promising technologies for the future generation wireless networks because of its advantages of providing a ultra-wide new spectrum and ultra-high data transmission rate. To reduce the power consumption and circuit cost for mmWave systems, hybrid digital and analog (HDA) architecture is preferred in such a scenario. In this paper, an artificial-noise (AN) aided secure HDA beamforming scheme is proposed for mmWave MISO system with low resolution digital-to-analog converters (DACs) and finite-quantized phase shifters on RF. The additive quantization noise model for AN aided HDA system is established to make an analysis of the secrecy performance of such systems. With the partial channel knowledge of eavesdropper available, an approximate expression of secrecy rate (SR) is derived. Then using this approximation formula, we propose a two-layer alternately iterative structure (TLAIS) for optimizing digital precoder (DP) of confidential message (CM), digital AN projection matrix (DANPM) and analog precoder (AP). The inner-layer iteration loop is to design the DP of CMs and DANPM alternatively given a fixed matrix of AP. The outer-layer iteration loop is in between digital baseband part and analog part, where the former refers to DP and DANPM, and the latter is AP. Then for a given digital part, we propose a gradient ascent algorithm to find the vector of AP vector. Given a matrix of AP, we make use of general power iteration (GPI) method to compute DP and DANPM. This process is repeated until the terminal condition is reached. Simulation results show that the proposed TLAIS can achieve a better SR performance compared to existing methods, especially in the high signal-to-noise ratio region.

cs.IT↗
arXiv2019

Performance Analysis of Directional Modulation with Finite-quantized RF Phase Shifters in Analog Beamforming Structure

The radio frequency (RF) phase shifter with finite quantization bits in analog beamforming (AB) structure forms quantization error (QE) and causes a performance loss of received signal to interference plus noise ratio (SINR) at the receiver (called Bob). By using the law of large numbers in probability theory, the closed-form expression of SINR performance loss is derived to be inversely proportional to the square of sinc (or sin(x)/x) function. Here, a phase alignment method is applied in directional modulation transmitter with AB structure. Also, the secrecy rate (SR) expression is derived with QE. From numerical simulation results, we find that the SINR performance loss gradually decreases as the number L of quantization bits increases. This loss is less than 0.3dB when L is larger than or equal to 3. As L exceeds 5, the SINR performance loss at Bob can be approximately trivial. Similarly, SR performance loss gradually reduces as L increases. In particular, the SR performance loss is about 0.1 bits/s/Hz for L=3 at signal-to-noise ratio of 15dB.

cs.IT↗
arXiv2019

Deep Octonion Networks

Deep learning is a research hot topic in the field of machine learning. Real-value neural networks (Real NNs), especially deep real networks (DRNs), have been widely used in many research fields. In recent years, the deep complex networks (DCNs) and the deep quaternion networks (DQNs) have attracted more and more attentions. The octonion algebra, which is an extension of complex algebra and quaternion algebra, can provide more efficient and compact expression. This paper constructs a general framework of deep octonion networks (DONs) and provides the main building blocks of DONs such as octonion convolution, octonion batch normalization and octonion weight initialization; DONs are then used in image classification tasks for CIFAR-10 and CIFAR-100 data sets. Compared with the DRNs, the DCNs, and the DQNs, the proposed DONs have better convergence and higher classification accuracy. The success of DONs is also explained by multi-task learning.

cs.CV↗
arXiv2018

Machine-Learning-based High-resolution DOA Measurement and Robust DM for Hybrid Analog-Digital Massive MIMO Transceiver

At hybrid analog-digital (HAD) transceiver, an improved HAD rotational invariance techniques (ESPRIT), called I-HAD-ESPRIT, is proposed to measure the direction of arrival (DOA) of desired user, where the phase ambiguity due to HAD structure is addressed successfully. Subsequently, a machine-learning (ML) framework is proposed to improve the precision of measuring DOA. In the training stage, the HAD transceiver works as a receiver and repeatedly measures the values of DOA via I-HAD-ESPRIT to form a slightly large training data set (TDS) of DOA. From TDS, we find that the probability density function (PDF) of DOA measurement error (DOAME) is approximated as a Gaussian distribution by the histogram method in ML. This TDS is used to learn the mean of DOA and the variance of DOAME, which are utilized to infer their values and improve their precisions in the real-time stage. The HAD transceiver rapidly measures the real-time value of DOA some times to form a relatively small real-time set (RTS), which is used to learn the real-time mean and variance of DOA/ DOAME. Then, three weight combiners are proposed to combine the-maximum-likelihood-learning outputs of TDS and RTS. Their weight factors depend intimately on the numbers of elements in TDS and RTS, and signal-to-noise ratios during the two stages. Using the mean and variance of DOA/DOAME, their PDFs can be given directly, and we propose a robust beamformer for directional modulation (DM) transmitter with HAD by fully exploiting the PDF of DOA/DOAME, especially a robust analog beamformer on RF chain. Simulation results show that: 1) The proposed I-HAD-ESPRIT can achieve the HAD CRLB; 2) The proposed ML framework performs much better than the corresponding real-time one without training stage, 3) The proposed robust DM transmitter can perform better than the corresponding non-robust ones in terms of bit error rate and secrecy rate.

cs.IT↗
arXiv2018

Covert Communication with A Full-Duplex Receiver Based on Channel Distribution Information

In this work, we consider a system of covert communication with the aid of a full-duplex (FD) receiver to enhance the performance in a more realistic scenario, i.e., only the channel distribution information (CDI) rather than channel state information (CSI) is known to a warden. Our work shows that transmitting random AN can improve the covert communication with the infinite blocklength. Specifically, we jointly design the optimal transmit power and AN power by minimizing the outage probability at Bob, and we find that the outage probability decreases and then increases as the maximum allowable AN power increases. Intuitively, once AN exceeds an optimal value, the performance will become worse because of the self-interference. The simulation results also show that the performance behaviors of CDI and CSI are different. When Willie only knows CDI, there is an optimal AN power that minimizes Bob's outage probability. However, when Willie knows CSI, the outage probability monotonically decreases with AN power.

eess.SP↗
arXiv2018

Directional Modulation: A Secure Solution to 5G and Beyond Mobile Networks

Directional modulation (DM), as an efficient secure transmission way, offers security through its directive property and is suitable for line-of-propagation (LoP) channels such as millimeter wave (mmWave) massive multiple-input multiple-output (MIMO), satellite communication, unmanned aerial vehicle (UAV), and smart transportation. If the direction angle of the desired received is known, the desired channel gain vector is obtainable. Thus, in advance, the DM transmitter knows the values of directional angles of desired user and eavesdropper, or their direction of arrival (DOAs) because the beamforming vector of confidential messages and artificial noise (AN) projection matrix is mainly determined by directional angles of desired user and eavesdropper. For a DM transceiver, working as a receiver, the first step is to measure the DOAs of desired user and eavesdropper. Then, in the second step, using the measured DOAs, the beamforming vector of confidential messages and AN projection matrix is designed. In this paper, we describe the DOA measurement methods, power allocation, and beamforming in DM networks. A machine learning-based DOA measurement method is proposed to make a substantial SR performance gain compared to single-snapshot measurement without machine learning for a given null-space projection beamforming scheme. However, for a conventional DM network, there still exists a serious secure issue: the eavesdropper moves inside the main beam of the desired user and may intercept the confidential messages intended to the desired users because the beamforming vector of confidential messages and AN projection matrix are only angle-dependence. To address this problem, we present a new concept of secure and precise transmission, where the transmit waveform has two-dimensional even three-dimensional dependence by using DM, random frequency selection, and phase alignment at DM transmitter.

cs.IT↗
arXiv2018

GPI-based Secrecy Rate Maximization Beamforming Scheme for Wireless Transmission with AN-aided Directional Modulation

In a directional modulation network, a general power iterative (GPI) based beamforming scheme is proposed to maximize the secrecy rate (SR), where there are two optimization variables required to be optimized. The first one is the useful precoding vector of transmitting confidential messages to the desired user while the second one is the artificial noise (AN) projection matrix of forcing more AN to eavesdroppers. In such a secure network, the paramount problem is how to design or optimize the two optimization variables by different criteria. To maximize the SR (Max-SR), an alternatively iterative structure (AIS) is established between the AN projection matrix and the precoding vector for confidential messages. To choose a good initial value of iteration process of GPI, the proposed Max-SR method can readily double its convergence speed compared to the random choice of initial value. With only four iterations, it may rapidly converge to its rate ceil. From simulation results, it follows that the SR performance of the proposed AIS of GPI-based Max-SR is much better than those of conventional leakage-based and null-space projection methods in the medium and large signal-to-noise ratio (SNR) regions, and its achievable SR performance gain gradually increases as SNR increases.

cs.IT↗
arXiv2017

Artificial-noise-aided Secure Multicast Precoding for Directional Modulation Systems

In multi-cast scenario, all desired users are divided into $K$ groups. Each group receives its own individual confidential message stream. Eavesdropper group aims to intercept $K$ confidential-message streams. To achieve a secure transmission, two secure schemes are proposed: maximum group receive power plus null-space (NS) projection (Max-GRP plus NSP) and leakage. The former obtains its precoding vector per group by maximizing its own group receive power subject to the orthogonal constraint, and its AN projection matrix consist of all bases of NS of all desired steering vectors from all groups. The latter attains its desired precoding vector per group by driving the current confidential message power to its group steering space and reducing its power leakage to eavesdropper group and other $K-1$ desired ones by maximizing signal to leakage and noise ratio (Max-SLNR). And its AN projection matrix is designed by forcing AN power into the eavesdropper steering space by viewing AN as a useful signal for eavesdropper group and maximizing AN to leakage-and-noise ratio (Max-ANLNR). Simulation results show that the proposed two methods are better than conventional method in terms of both bit-error-rate (BER) and secrecy sum-rate per group. Also, the leakage scheme performs better than Max-GRP-NSP , especially in the presence of direction measurement errors. However, the latter requires no channel statistical parameters and thus is simpler compared to the former.

cs.IT↗
arXiv2014

Start-up vortex flow past an accelerated flat plate

Viscous flow past a finite flat plate moving in direction normal to itself is studied numerically.The plate moves with velocity $at^p$, where $p=0,0.5,1,2$. We present the evolution of vorticity profiles, streaklines and streamlines, and study the dependence on the acceleration parameter $p$. Four stages in the vortex evolution, as proposed by Luchini & Tognaccini (2002), are clearly identified. The initial stage, in which the vorticity consists solely of a Rayleigh boundary layer, is shown to last for a time-interval whose length shrinks to zero like $p^3$, as $p \to 0$. In the second stage, a center of rotation develops near the tip of the plate, well before a vorticity maximum within the vortex core develops. Once the vorticity maximum develops, its position oscillates and differs from the center of rotation. The difference between the two increases with increasing $p$, and decreases in time. In the third stage, the center of rotation and the shed circulation closely satisfy self-similar scaling laws for inviscid flow. Finally, in the fourth stage, the finite plate length becomes relevant and the flow begins to depart from the self-similar behaviour. While the core trajectory and circulation closely satisfy inviscid scaling laws, the vorticity maximum and the boundary layer thickness follow viscous scaling laws. The results are compared with experimental results of Pullin & Perry (1980), and Taneda & Honji (1971), where available.

physics.flu-dyn↗
arXiv2014

Numerical study of viscous starting flow past a flat plate

Viscous flow past a finite plate which is impulsively started in direction normal to itself is studied numerically using a high order mixed finite difference and semi-Lagrangian scheme. The goal is to resolve details of the vorticity generation at early times, and to determine the effect of viscosity on flow quantities such as the core trajectory and vorticity, and the shed circulation. Vorticity contours, streaklines and streamlines are presented for a range of Reynolds numbers $Re \in [250, 2000]$ and a range of times $t \in[0. 0002, 5]$. At early times, most of the vorticity is attached to the plate. The paper proposes a definition for the shed circulation at early as well as late times, and shows that it indeed represents vorticity that separates from the plate without reattaching. The contribution of viscous diffusion to the circulation shedding rate is found to be significant, but, interestingly, to depend only slightly on the value of the Reynolds number. The shed circulation and the vortex core trajectories follow scaling laws for inviscid self-similar flow over several decades in time. Scaling laws describing the core vorticity, core dissipation, boundary layer thickness, drag and lift forces in time and Reynolds number are also presented. The simulations provide benchmark results to evaluate, for example, simpler separation models such as point vortex and vortex sheet models.

physics.flu-dyn↗
arXiv2013

On Galerkin Approximations for the Zakai Equation with Diffusive and Point Process Observations

This paper studies Galerkin approximations applied to the Zakai equation of stochastic filtering. The basic idea of this approach is to project the infinite-dimensional Zakai equation onto some finite-dimensional subspace generated by smooth basis functions; this leads to a finite-dimensional system of stochastic differential equations that can be solved numerically. The contribution of the paper is twofold. On the theoretical side, existing convergence results are extended to filtering models with observations of point-process or mixed type. On the applied side, various issues related to the numerical implementation of the method are considered; in particular, we propose to work with a subspace that is constructed from a basis of Hermite polynomials. The paper closes with a numerical case study.

math.NA↗
arXiv2012

Variational formulas of higher order mean curvatures

In this paper, we establish the first variational formula and its Euler-Lagrange equation for the total $2p$-th mean curvature functional $\mathcal {M}_{2p}$ of a submanifold $M^n$ in a general Riemannian manifold $N^{n+m}$ for $p=0,1,...,[\frac{n}{2}]$. As an example, we prove that closed complex submanifolds in complex projective spaces are critical points of the functional $\mathcal {M}_{2p}$, called relatively $2p$-minimal submanifolds, for all $p$. At last, we discuss the relations between relatively $2p$-minimal submanifolds and austere submanifolds in real space forms, as well as a special variational problem.

math.DG↗