arXiv subjects

Yong Xu

Publications and source records attributed to Yong Xu.

At least 361 records · Page 20Linked to original sources

arXiv2020

Sound Event Detection of Weakly Labelled Data with CNN-Transformer and Automatic Threshold Optimization

Sound event detection (SED) is a task to detect sound events in an audio recording. One challenge of the SED task is that many datasets such as the Detection and Classification of Acoustic Scenes and Events (DCASE) datasets are weakly labelled. That is, there are only audio tags for each audio clip without the onset and offset times of sound events. \qk{We compare segment-wise and clip-wise training for SED that is lacking in previous works. We propose a convolutional neural network transformer (CNN-Transfomer) for audio tagging and SED, and show that CNN-Transformer performs similarly to a convolutional recurrent neural network (CRNN)}. Another challenge of SED is that thresholds are required for detecting sound events. Previous works set thresholds empirically, and are not an optimal approaches. To solve this problem, we propose an automatic threshold optimization method. The first stage is to optimize the system with respect to metrics that do not depend on thresholds, such as mean average precision (mAP). The second stage is to optimize the thresholds with respect to metrics that depends on those thresholds. Our proposed automatic threshold optimization system achieves a state-of-the-art audio tagging F1 of 0.646, outperforming that without threshold optimization of 0.629, and a sound event detection F1 of 0.584, outperforming that without threshold optimization of 0.564.

cs.SD↗

arXiv2020

Deep Learning on Image Denoising: An overview

Deep learning techniques have received much attention in the area of image denoising. However, there are substantial differences in the various types of deep learning methods dealing with image denoising. Specifically, discriminative learning based on deep learning can ably address the issue of Gaussian noise. Optimization models based on deep learning are effective in estimating the real noise. However, there has thus far been little related research to summarize the different deep learning techniques for image denoising. In this paper, we offer a comparative study of deep techniques in image denoising. We first classify the deep convolutional neural networks (CNNs) for additive white noisy images; the deep CNNs for real noisy images; the deep CNNs for blind denoising and the deep CNNs for hybrid noisy images, which represents the combination of noisy, blurred and low-resolution images. Then, we analyze the motivations and principles of the different types of deep learning methods. Next, we compare the state-of-the-art methods on public denoising datasets in terms of quantitative and qualitative analysis. Finally, we point out some potential challenges and directions of future research.

eess.IV↗

arXiv2020

Neural Spatio-Temporal Beamformer for Target Speech Separation

Purely neural network (NN) based speech separation and enhancement methods, although can achieve good objective scores, inevitably cause nonlinear speech distortions that are harmful for the automatic speech recognition (ASR). On the other hand, the minimum variance distortionless response (MVDR) beamformer with NN-predicted masks, although can significantly reduce speech distortions, has limited noise reduction capability. In this paper, we propose a multi-tap MVDR beamformer with complex-valued masks for speech separation and enhancement. Compared to the state-of-the-art NN-mask based MVDR beamformer, the multi-tap MVDR beamformer exploits the inter-frame correlation in addition to the inter-microphone correlation that is already utilized in prior arts. Further improvements include the replacement of the real-valued masks with the complex-valued masks and the joint training of the complex-mask NN. The evaluation on our multi-modal multi-channel target speech separation and enhancement platform demonstrates that our proposed multi-tap MVDR beamformer improves both the ASR accuracy and the perceptual speech quality against prior arts.

eess.AS↗

arXiv2020

High-Temperature Quantum Anomalous Hall Insulators in Lithium-Decorated Iron-Based Superconductor Materials

Quantum anomalous Hall (QAH) insulator is the key material to study emergent topological quantum effects, but its ultralow working temperature limits experiments. Here, by first-principles calculations, we find a family of stable two-dimensional (2D) structures generated by lithium decoration of layered iron-based superconductor materials FeX (X = S, Se, Te), and predict room-temperature ferromagnetic semiconductors together with large-gap high-Chern-number QAH insulators in the 2D materials. The extremely robust ferromagnetic order is induced by the electron injection from Li to Fe and stabilized by strong ferromagnetic kinetic exchange in the 2D Fe layer. While in the absence of spin-orbit coupling (SOC), the ferromagnetism polarizes the system into a half Dirac semimetal state protected by mirror symmetry, the SOC effect results in a spontaneous breaking of mirror symmetry and introduces a Dirac mass term, which creates QAH states with sizable gaps (several tens of meV) and multiple chiral edge modes. We also find a 3D QAH insulator phase featured by macroscopic number of chiral conduction channels in bulk LiOH-LiFeX. The findings open new opportunities to realize novel QAH physics and applications at high temperatures.

cond-mat.mtrl-sci↗

arXiv2020

Hidden Physical Effects in Non-centrosymmetric Crystals

Symmetry forbidden effects in crystals may emerge in a local environment that breaks the symmetries. Yet these hidden physical effects were only discussed in centrosymmetric crystals. Here we propose that hidden physical effects can be generalized to almost all crystallographic symmetric groups and hence reveal their universality and diversity. We systematically discuss certain symmetries in crystals that may induce hidden spin polarization (HSP), hidden berry curvature and hidden valley polarization, with a focus on a specific pattern of HSP whose winding directions of the in-plane spin vectors are the same for adjacent bands, dubbed anomalous HSP. Such an unprecedented spin pattern arises from relatively weak spin-dependent inter-sector interaction and is demonstrated in mirror-symmetric InSe by first-principles calculations. Unique electric field-dependent splitting into specific spatial polarization pattern may serve as an experimental signature of anomalous HSP. Our results reveal abundant hidden physical effects beyond centrosymmetric crystals and provide new platforms to discuss them for emergent physical effects and future applications.

cond-mat.mtrl-sci↗

arXiv2020

Lightweight image super-resolution with enhanced CNN

Deep convolutional neural networks (CNNs) with strong expressive ability have achieved impressive performances on single image super-resolution (SISR). However, their excessive amounts of convolutions and parameters usually consume high computational cost and more memory storage for training a SR model, which limits their applications to SR with resource-constrained devices in real world. To resolve these problems, we propose a lightweight enhanced SR CNN (LESRCNN) with three successive sub-blocks, an information extraction and enhancement block (IEEB), a reconstruction block (RB) and an information refinement block (IRB). Specifically, the IEEB extracts hierarchical low-resolution (LR) features and aggregates the obtained features step-by-step to increase the memory ability of the shallow layers on deep layers for SISR. To remove redundant information obtained, a heterogeneous architecture is adopted in the IEEB. After that, the RB converts low-frequency features into high-frequency features by fusing global and local features, which is complementary with the IEEB in tackling the long-term dependency problem. Finally, the IRB uses coarse high-frequency features from the RB to learn more accurate SR features and construct a SR image. The proposed LESRCNN can obtain a high-quality image by a model for different scales. Extensive experiments demonstrate that the proposed LESRCNN outperforms state-of-the-arts on SISR in terms of qualitative and quantitative evaluation. The code of LESRCNN is accessible on https://github.com/hellloxiaotian/LESRCNN.

eess.IV↗

arXiv2020

Recurrent Exposure Generation for Low-Light Face Detection

Face detection from low-light images is challenging due to limited photos and inevitable noise, which, to make the task even harder, are often spatially unevenly distributed. A natural solution is to borrow the idea from multi-exposure, which captures multiple shots to obtain well-exposed images under challenging conditions. High-quality implementation/approximation of multi-exposure from a single image is however nontrivial. Fortunately, as shown in this paper, neither is such high-quality necessary since our task is face detection rather than image enhancement. Specifically, we propose a novel Recurrent Exposure Generation (REG) module and couple it seamlessly with a Multi-Exposure Detection (MED) module, and thus significantly improve face detection performance by effectively inhibiting non-uniform illumination and noise issues. REG produces progressively and efficiently intermediate images corresponding to various exposure settings, and such pseudo-exposures are then fused by MED to detect faces across different lighting conditions. The proposed method, named REGDet, is the first `detection-with-enhancement' framework for low-light face detection. It not only encourages rich interaction and feature fusion across different illumination levels, but also enables effective end-to-end learning of the REG component to be better tailored for face detection. Moreover, as clearly shown in our experiments, REG can be flexibly coupled with different face detectors without extra low/normal-light image pairs for training. We tested REGDet on the DARK FACE low-light face benchmark with thorough ablation study, where REGDet outperforms previous state-of-the-arts by a significant margin, with only negligible extra parameters.

cs.CV↗

arXiv2020

Averaging principles for non-autonomous two-time-scale stochastic reaction-diffusion equations with jump

In this paper, we aim to develop the averaging principle for a slow-fast system of stochastic reaction-diffusion equations driven by Poisson random measures. The coefficients of the equation are assumed to be functions of time, and some of them are periodic or almost periodic. Therefore, the Poisson term needs to be processed, and a new averaged equation needs to be given. For this reason, the existence of time-dependent evolution family of measures associated with the fast equation is studied, and proved that it is almost periodic. Next, according to the characteristics of almost periodic functions, the averaged coefficient is defined by the evolution family of measures, and the averaged equation is given. Finally, the validity of the averaging principle is verified by using the Khasminskii method.

math.DS↗

arXiv2020

Winding Numbers and Generalized Mobility Edges in Non-Hermitian Systems

The Aubry-André-Harper (AAH) model with a self-dual symmetry plays an important role in studying the Anderson localization. Here we find a self-dual symmetry determining the quantum phase transition between extended and localized states in a non-Hermitian AAH model and show that the eigenenergies of these states are characterized by two types of winding numbers. By constructing and studying a non-Hermitian generalized AAH model, we further generalize the notion of the mobility edge, which separates the localized and extended states in the energy spectrum of disordered systems, to the non-Hermitian case and find that the generalized mobility edge is of a topological nature even in the open boundary geometry in the sense that the energies of localized and extended states exhibit distinct topological structures in the complex energy plane. Finally, we propose an experimental scheme to realize these models with electric circuits.

cond-mat.mes-hall↗

arXiv2020

Designing and Training of A Dual CNN for Image Denoising

Deep convolutional neural networks (CNNs) for image denoising have recently attracted increasing research interest. However, plain networks cannot recover fine details for a complex task, such as real noisy images. In this paper, we propsoed a Dual denoising Network (DudeNet) to recover a clean image. Specifically, DudeNet consists of four modules: a feature extraction block, an enhancement block, a compression block, and a reconstruction block. The feature extraction block with a sparse machanism extracts global and local features via two sub-networks. The enhancement block gathers and fuses the global and local features to provide complementary information for the latter network. The compression block refines the extracted information and compresses the network. Finally, the reconstruction block is utilized to reconstruct a denoised image. The DudeNet has the following advantages: (1) The dual networks with a parse mechanism can extract complementary features to enhance the generalized ability of denoiser. (2) Fusing global and local features can extract salient features to recover fine details for complex noisy images. (3) A Small-size filter is used to reduce the complexity of denoiser. Extensive experiments demonstrate the superiority of DudeNet over existing current state-of-the-art denoising methods.

eess.IV↗

arXiv2020

Type-II quadrupole topological insulators

Modern theory of electric polarization is formulated by the Berry phase, which, when quantized, leads to topological phases of matter. Such a formulation has recently been extended to higher electric multipole moments, through the discovery of the so-called quadupole topological insulator. It has been established by a classical electromagnetic theory that in a two-dimensional material the quantized properties for the quadupole topological insulator should satisfy a basic relation. Here we discover a new type of quadrupole topological insulator (dubbed type-II) that violates this relation due to the breakdown of the correspondence that a Wannier band and an edge energy spectrum close their gaps simultaneously. We find that, similar to the previously discovered (referred to as type-I) quadrupole topological insulator, the type-II hosts topologically protected corner states carrying fractional corner charges. However, the edge polarizations only occur at a pair of boundaries in the type-II insulating phase, leading to the violation of the classical constraint. We demonstrate that such new topological phenomena can appear from quench dynamics in non-equilibrium systems, which can be experimentally observed in ultracold atomic gases. We also propose an experimental scheme with electric circuits to realize such a new topological phase of matter. The existence of the new topological insulating phase means that new multipole topological insulators with distinct properties can exist in broader contexts beyond classical constraints.

cond-mat.mes-hall↗

arXiv2020

Deep Bilateral Retinex for Low-Light Image Enhancement

Low-light images, i.e. the images captured in low-light conditions, suffer from very poor visibility caused by low contrast, color distortion and significant measurement noise. Low-light image enhancement is about improving the visibility of low-light images. As the measurement noise in low-light images is usually significant yet complex with spatially-varying characteristic, how to handle the noise effectively is an important yet challenging problem in low-light image enhancement. Based on the Retinex decomposition of natural images, this paper proposes a deep learning method for low-light image enhancement with a particular focus on handling the measurement noise. The basic idea is to train a neural network to generate a set of pixel-wise operators for simultaneously predicting the noise and the illumination layer, where the operators are defined in the bilateral space. Such an integrated approach allows us to have an accurate prediction of the reflectance layer in the presence of significant spatially-varying measurement noise. Extensive experiments on several benchmark datasets have shown that the proposed method is very competitive to the state-of-the-art methods, and has significant advantage over others when processing images captured in extremely low lighting conditions.

eess.IV↗

arXiv2020

Distortionless Multi-Channel Target Speech Enhancement for Overlapped Speech Recognition

Speech enhancement techniques based on deep learning have brought significant improvement on speech quality and intelligibility. Nevertheless, a large gain in speech quality measured by objective metrics, such as perceptual evaluation of speech quality (PESQ), does not necessarily lead to improved speech recognition performance due to speech distortion in the enhancement stage. In this paper, a multi-channel dilated convolutional network based frequency domain modeling is presented to enhance target speaker in the far-field, noisy and multi-talker conditions. We study three approaches towards distortionless waveforms for overlapped speech recognition: estimating complex ideal ratio mask with an infinite range, incorporating the fbank loss in a multi-objective learning and finetuning the enhancement model by an acoustic model. Experimental results proved the effectiveness of all three approaches on reducing speech distortions and improving recognition accuracy. Particularly, the jointly tuned enhancement model works very well with other standalone acoustic model on real test data.

eess.AS↗

arXiv2020

Type-II Ising superconductivity and anomalous metallic state in macro-size ambient-stable ultrathin crystalline films

Recent emergence of two-dimensional (2D) crystalline superconductors has provided a promising platform to investigate novel quantum physics and potential applications. To reveal essential quantum phenomena therein, ultralow temperature transport investigation on high quality ultrathin superconducting films is critically required, although it has been quite challenging experimentally. Here we report a systematic transport study on the ultrathin crystalline PdTe2 films grown by molecular beam epitaxy (MBE). Interestingly, a new type of Ising superconductivity in 2D centrosymmetric materials is revealed by the detection of large in-plane critical field more than 7 times Pauli limit. Remarkably, in perpendicular magnetic field, we provide solid evidence of anomalous metallic state characterized by the resistance saturation at low temperatures with high quality filters. The robust superconductivity with intriguing quantum phenomena in the macro-size ambient-stable ultrathin PdTe2 films remains almost the same for 20 months, showing great potentials in electronic and spintronic applications.

cond-mat.supr-con↗

arXiv2020

Higher-order topological insulators and semimetals in generalized Aubry-André-Harper models

Higher-order topological phases of matter have been extensively studied in various areas of physics. While the Aubry-André-Harper model provides a paradigmatic example to study topological phases, it has not been explored whether a generalized Aubry-André-Harper model can exhibit a higher-order topological phenomenon. Here, we construct a two-dimensional higher-order topological insulator with chiral symmetry based on the Aubry-André-Harper model. We find the coexistence of zero-energy and nonzero energy corner-localized modes. The former is protected by the quantized quadrupole moment, while the latter by the first Chern number of the Wannier band. The nonzero-energy mode can also be viewed as the consequence of a Chern insulator localized on a surface. More interestingly, the non-zero energy corner mode can lie in the continuum of extended bulk states and form a bound state in the continuum of higher-order topological systems. We finally propose an experimental scheme to realize our model in electric circuits. Our study opens a door to further study higher-order topological phases based on the Aubry-André-Harper model.

cond-mat.mes-hall↗

arXiv2020

High-Chern-Number and High-Temperature Quantum Hall Effect without Landau Levels

The quantum Hall effect (QHE) with quantized Hall resistance of h/νe2 starts the research on topological quantum states and lays the foundation of topology in physics. Afterwards, Haldane proposed the QHE without Landau levels, showing nonzero Chern number |C|=1, which has been experimentally observed at relatively low temperatures. For emerging physics and low-power-consumption electronics, the key issues are how to increase the working temperature and realize high Chern numbers (C>1). Here, we report the experimental discovery of high-Chern-number QHE (C=2) without Landau levels and C=1 Chern insulator state displaying nearly quantized Hall resistance plateau above the Néel temperature in MnBi2Te4 devices. Our observations provide a new perspective on topological matter and open new avenues for exploration of exotic topological quantum states and topological phase transitions at higher temperatures.

cond-mat.mes-hall↗

arXiv2020

Topological Insulators beyond Energy Band Characterization

Topological phases of matter are generally characterized by topological properties of energy bands of a system. Their transitions under preserved symmetries occur through closing a gap of energy bands, leading to topologically protected edge states in energy spectra in topological phases. Here we predict a new topological phase that emerges through closing a gap of bands constructed by energy bands, instead of through closing an energy gap with preserved symmetries. From this perspective, topological phases may arise from topological properties of the "bands of bands" associated with their gap closure and corresponding edge states. We demonstrate this idea by studying a tight-binding model. We find that the Wannier bands constructed by energy bands exhibit a gap closure associated with a change of a winding number, while the energy bands remain gapped and trivial without any zero energy modes. In addition, the topological Wannier bands give rise to quantized edge polarizations. Since the emergence of this topological phase does not involve any energy gap closure, we expect its appearance under unitary time evolution. Indeed, this phase appears as we perform a quench dynamics. Our study opens a new direction for exploring topological phases beyond conventional energy band characterization.

cond-mat.mes-hall↗

arXiv2020

Audio-Visual Speech Separation and Dereverberation with a Two-Stage Multimodal Network

Background noise, interfering speech and room reverberation frequently distort target speech in real listening environments. In this study, we address joint speech separation and dereverberation, which aims to separate target speech from background noise, interfering speech and room reverberation. In order to tackle this fundamentally difficult problem, we propose a novel multimodal network that exploits both audio and visual signals. The proposed network architecture adopts a two-stage strategy, where a separation module is employed to attenuate background noise and interfering speech in the first stage and a dereverberation module to suppress room reverberation in the second stage. The two modules are first trained separately, and then integrated for joint training, which is based on a new multi-objective loss function. Our experimental results show that the proposed multimodal network yields consistently better objective intelligibility and perceptual quality than several one-stage and two-stage baselines. We find that our network achieves a 21.10% improvement in ESTOI and a 0.79 improvement in PESQ over the unprocessed mixtures. Moreover, our network architecture does not require the knowledge of the number of speakers.

eess.AS↗