arXiv subjects

Gang Wang

Publications and source records attributed to Gang Wang.

At least 451 records · Page 25Linked to original sources

arXiv2016

Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos

Single modality action recognition on RGB or depth sequences has been extensively explored recently. It is generally accepted that each of these two modalities has different strengths and limitations for the task of action recognition. Therefore, analysis of the RGB+D videos can help us to better study the complementary properties of these two types of modalities and achieve higher levels of performance. In this paper, we propose a new deep autoencoder based shared-specific feature factorization network to separate input multimodal signals into a hierarchy of components. Further, based on the structure of the features, a structured sparsity learning machine is proposed which utilizes mixed norms to apply regularization within components and group selection between them for better classification performance. Our experimental results show the effectiveness of our cross-modality feature analysis framework by achieving state-of-the-art accuracy for action classification on five challenging benchmark datasets.

cs.CV↗

arXiv2016

Multi-resonant piezoelectric shunting induced by digital controllers for subwavelength elastic wave attenuation in smart metamaterial

Instead of analog electronic circuits and components, digital controllers that are capable of active multi-resonant piezoelectric shunting are applied to elastic metamaterials integrated with piezoelectric patches. Giving thanks to the introduced digital control technique, shunting strategies with transfer functions that can hardly be realized with analog circuits is possible now. As an example, the "pole-zero" method is developed to design single- or multi-resonant bandgaps by adjusting poles and zeros in the transfer function of piezoelectric shunting directly. Large simultaneous attenuations in up to three frequency bands at deep subwavelength scale (with the normalized frequency as low as 0.077) are achieved. The underlying physical mechanism is attributed to the negative group velocity of flexural wave within bandgaps. As digital controllers can be readily adapted via wireless broadcasting, the bandgaps can be tuned easily instead of tuning the electric components in analog shunting circuits one by one manually. The theoretical results are well verified experimentally with the measured vibration transmission properties where large insulations of up to 20dB in low-frequency ranges are observed.

cond-mat.mtrl-sci↗

arXiv2016

Improving Fully Convolution Network for Semantic Segmentation

Fully Convolution Networks (FCN) have achieved great success in dense prediction tasks including semantic segmentation. In this paper, we start from discussing FCN by understanding its architecture limitations in building a strong segmentation network. Next, we present our Improved Fully Convolution Network (IFCN). In contrast to FCN, IFCN introduces a context network that progressively expands the receptive fields of feature maps. In addition, dense skip connections are added so that the context network can be effectively optimized. More importantly, these dense skip connections enable IFCN to fuse rich-scale context to make reliable predictions. Empirically, those architecture modifications are proven to be significant to enhance the segmentation performance. Without engaging any contextual post-processing, IFCN significantly advances the state-of-the-arts on ADE20K (ImageNet scene parsing), Pascal Context, Pascal VOC 2012 and SUN-RGBD segmentation datasets.

cs.CV↗

arXiv2016

Solving Large-scale Systems of Random Quadratic Equations via Stochastic Truncated Amplitude Flow

A novel approach termed \emph{stochastic truncated amplitude flow} (STAF) is developed to reconstruct an unknown $n$-dimensional real-/complex-valued signal $\bm{x}$ from $m$ `phaseless' quadratic equations of the form $ψ_i=|\langle\bm{a}_i,\bm{x}\rangle|$. This problem, also known as phase retrieval from magnitude-only information, is \emph{NP-hard} in general. Adopting an amplitude-based nonconvex formulation, STAF leads to an iterative solver comprising two stages: s1) Orthogonality-promoting initialization through a stochastic variance reduced gradient algorithm; and, s2) A series of iterative refinements of the initialization using stochastic truncated gradient iterations. Both stages involve a single equation per iteration, thus rendering STAF a simple, scalable, and fast approach amenable to large-scale implementations that is useful when $n$ is large. When $\{\bm{a}_i\}_{i=1}^m$ are independent Gaussian, STAF provably recovers exactly any $\bm{x}\in\mathbb{R}^n$ exponentially fast based on order of $n$ quadratic equations. STAF is also robust in the presence of additive noise of bounded support. Simulated tests involving real Gaussian $\{\bm{a}_i\}$ vectors demonstrate that STAF empirically reconstructs any $\bm{x}\in\mathbb{R}^n$ exactly from about $2.3n$ magnitude-only measurements, outperforming state-of-the-art approaches and narrowing the gap from the information-theoretic number of equations $m=2n-1$. Extensive experiments using synthetic data and real images corroborate markedly improved performance of STAF over existing alternatives.

cs.IT↗

arXiv2016

Decision Making Based on Cohort Scores for Speaker Verification

Decision making is an important component in a speaker verification system. For the conventional GMM-UBM architecture, the decision is usually conducted based on the log likelihood ratio of the test utterance against the GMM of the claimed speaker and the UBM. This single-score decision is simple but tends to be sensitive to the complex variations in speech signals (e.g. text content, channel, speaking style, etc.). In this paper, we propose a decision making approach based on multiple scores derived from a set of cohort GMMs (cohort scores). Importantly, these cohort scores are not simply averaged as in conventional cohort methods; instead, we employ a powerful discriminative model as the decision maker. Experimental results show that the proposed method delivers substantial performance improvement over the baseline system, especially when a deep neural network (DNN) is used as the decision maker, and the DNN input involves some statistical features derived from the cohort scores.

cs.SD↗

arXiv2016

Gated Siamese Convolutional Neural Network Architecture for Human Re-Identification

Matching pedestrians across multiple camera views, known as human re-identification, is a challenging research problem that has numerous applications in visual surveillance. With the resurgence of Convolutional Neural Networks (CNNs), several end-to-end deep Siamese CNN architectures have been proposed for human re-identification with the objective of projecting the images of similar pairs (i.e. same identity) to be closer to each other and those of dissimilar pairs to be distant from each other. However, current networks extract fixed representations for each image regardless of other images which are paired with it and the comparison with other images is done only at the final level. In this setting, the network is at risk of failing to extract finer local patterns that may be essential to distinguish positive pairs from hard negative pairs. In this paper, we propose a gating function to selectively emphasize such fine common local patterns by comparing the mid-level features across pairs of images. This produces flexible representations for the same image according to the images they are paired with. We conduct experiments on the CUHK03, Market-1501 and VIPeR datasets and demonstrate improved performance compared to a baseline Siamese CNN architecture.

cs.CV↗

arXiv2016

Joint Learning of Siamese CNNs and Temporally Constrained Metrics for Tracklet Association

In this paper, we study the challenging problem of multi-object tracking in a complex scene captured by a single camera. Different from the existing tracklet association-based tracking methods, we propose a novel and efficient way to obtain discriminative appearance-based tracklet affinity models. Our proposed method jointly learns the convolutional neural networks (CNNs) and temporally constrained metrics. In our method, a Siamese convolutional neural network (CNN) is first pre-trained on the auxiliary data. Then the Siamese CNN and temporally constrained metrics are jointly learned online to construct the appearance-based tracklet affinity models. The proposed method can jointly learn the hierarchical deep features and temporally constrained segment-wise metrics under a unified framework. For reliable association between tracklets, a novel loss function incorporating temporally constrained multi-task learning mechanism is proposed. By employing the proposed method, tracklet association can be accomplished even in challenging situations. Moreover, a new dataset with 40 fully annotated sequences is created to facilitate the tracking evaluation. Experimental results on five public datasets and the new large-scale dataset show that our method outperforms several state-of-the-art approaches in multi-object tracking.

cs.CV↗

arXiv2016

Decoupled Uplink and Downlink in a Wireless System with Buffer-Aided Relaying

The paper treats a multiuser relay scenario where multiple user equipments (UEs) have a two-way communication with a common Base Station (BS) in the presence of a buffer-equipped Relay Station (RS). Each of the uplink (UL) and downlink (DL) transmission can take place over a direct or over a relayed path. Traditionally, the UL and the DL path of a given two-way link are coupled, that is, either both are direct links or both are relayed links. By removing the restriction for coupling, one opens the design space for a decoupled two-way links. Following this, we devise two protocols: orthogonal decoupled UL/DL buffer-aided (ODBA) relaying protocol and non-orthogonal decoupled UL/DL buffer-aided (NODBA) relaying protocol. In NODBA, the receiver can use successive interference cancellation (SIC) to extract the desired signal from a collision between UL and DL signals. For both protocols, we characterize the transmission decision policies in terms of maximization of the average two-way sum rate of the system. The numerical results show that decoupling association and non-orthogonal radio access lead to significant throughput gains for two-way traffic.

cs.IT↗

arXiv2016

On Buffer-Aided Multiple-Access Relay Channel

The paper treats uplink scenario where M user equipments (UEs) send to a Base Station (BS), possibly via a common Relay Station (RS) that is equipped with a buffer. This is a multiple-access relay channel (MARC) aided by a buffer. We devise a protocol in which the transmission mode is selected adaptively, using the buffer at the RS in order to maximize the average system throughput. We consider the general case in which the RS and the BS can have limits on the maximal number of transmitters that can be received over the multiple access channel. In each slot there are three type possible actions: (A1) multiple UEs transmit at rates that enable BS to decode them (A2) multiple UEs transmit, the BS can only decode the messages partially, while the RS completely; (A3) RS forwards the side information to BS about the partially decoded messages, which are going to be combined and decoded entirely at the BS, while simultaneously a number of UEs sends new messages to the BS. The results show that the adaptive selection of direct and buffer-aided relay transmissions leads to significant average throughput gains.

cs.IT↗

arXiv2016

Symmetrical broken and nonlinear response of Weyl semimetal TaAs influenced by the topological surface states and Weyl nodes

A Weyl semimetal (WSM) features Weyl fermions in its bulk and topological surface states on surfaces, and is novel material hosting Weyl fermions, a kind of fundamental particles. The WSM was regarded as a three-dimensional version of "graphene" under the illusion. In order to explore its promising photoelectric properties and applications in photonics and photoelectronics, here, we study the anisotropic linear and nonlinear optical responses of a WSM TaAs, which are determined by the relationship and balance between its topological surface states and Weyl nodes. We demonstrate that topological surface states which break the bulk symmetry are responsible for the anisotropy of the mobility, and the anisotropic nonlinear response shows saturable characteristic with extremely large saturable intensity. We also find that the mobility is anisotropic with the magnitude of 104 cm2V-1s-1 at room temperature and can be accelerated by the optical field. By analyzing the symmetry, the nonlinear response is mainly contributed by the fermions close to the Weyl nodes, and is related to the Pauli's blocking of fermions, electron-electron interaction. This work experimentally discovers the anisotropic ultrahigh mobility of WSMs in the optical field and may start the field for the applications of WSMs in photonics and photoelectronics.

cond-mat.mtrl-sci↗

arXiv2016

Experimental results on chiral magnetic and vortical effects

Various novel transport phenomena in chiral systems result from the interplay of quantum anomalies with magnetic field and vorticity in high-energy heavy-ion collisions, and could survive the expansion of the fireball and be detected in experiments. Among them are the chiral magnetic effect, the chiral vortical effect and the chiral magnetic wave, the experimental searches for which have aroused extensive interest. The goal of this review is to describe the current status of experimental studies at Relativistic Heavy Ion Collider at BNL and the Large Hadron Collider at CERN, and to outline the future work in experiment needed to eliminate the existing uncertainties in the interpretation of the data.

nucl-ex↗

arXiv2016

Simultaneous Realization of Coherent Perfect Absorber and Laser by Zero-Index Media with both Gain and Loss

We investigate a unique type of zero-index medium with both gain and loss (ZIMGL), whose effective permittivity and permeability are purely imaginary and of opposite signs. We show analytically that, by using a slab of ZIM-GL with equal magnitude of loss and gain, coherent perfect absorber (CPA) and laser, i.e. the socalled CPA-laser, can be achieved simultaneously. CPA-laser has been found previously in parity-time (PT) symmetric systems. However, the underlying physics in a PT-symmetric system is distinct from that in a ZIM-GL. By designing a photonic crystal (PC) composed of core-shell rods, with loss and gain distributed in either the core or the shell, we have realized such a ZIM-GL. The CPA-laser functionality of such a PC is also confirmed in our numerical simulations. Our work provides a different approach for simultaneous realization of CPA and laser besides PTsymmetric systems.

physics.optics↗

arXiv2016

Chiral Magnetic Effect Task Force Report

In this report, we briefly examine the current status of the study of the chiral magnetic effect including theory and experimental progress. We recommend future strategies for resolving uncertainties in interpretation including recommendations for theoretical work, recommendations for measurements based on data collected in the past five years, and recommendations for beam use in the coming years of RHIC. We have specifically investigated the case for colliding nuclear isobars (nuclei with the same mass but different charge) and find the case compelling. We recommend that a program of nuclear isobar collisions to isolate the chiral magnetic effect from background sources be placed as a high priority item in the strategy for completing the RHIC mission.

nucl-th↗

arXiv2016

Learning Common and Specific Features for RGB-D Semantic Segmentation with Deconvolutional Networks

In this paper, we tackle the problem of RGB-D semantic segmentation of indoor images. We take advantage of deconvolutional networks which can predict pixel-wise class labels, and develop a new structure for deconvolution of multiple modalities. We propose a novel feature transformation network to bridge the convolutional networks and deconvolutional networks. In the feature transformation network, we correlate the two modalities by discovering common features between them, as well as characterize each modality by discovering modality specific features. With the common features, we not only closely correlate the two modalities, but also allow them to borrow features from each other to enhance the representation of shared information. With specific features, we capture the visual patterns that are only visible in one modality. The proposed network achieves competitive segmentation accuracy on NYU depth dataset V1 and V2.

cs.CV↗

arXiv2016

A Siamese Long Short-Term Memory Architecture for Human Re-Identification

Matching pedestrians across multiple camera views known as human re-identification (re-identification) is a challenging problem in visual surveillance. In the existing works concentrating on feature extraction, representations are formed locally and independent of other regions. We present a novel siamese Long Short-Term Memory (LSTM) architecture that can process image regions sequentially and enhance the discriminative capability of local feature representation by leveraging contextual information. The feedback connections and internal gating mechanism of the LSTM cells enable our model to memorize the spatial dependencies and selectively propagate relevant contextual information through the network. We demonstrate improved performance compared to the baseline algorithm with no LSTM units and promising results compared to state-of-the-art methods on Market-1501, CUHK03 and VIPeR datasets. Visualization of the internal mechanism of LSTM cells shows meaningful patterns can be learned by our method.

cs.CV↗

arXiv2016

Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition

3D action recognition - analysis of human actions based on 3D skeleton data - becomes popular recently due to its succinctness, robustness, and view-invariant representation. Recent attempts on this problem suggested to develop RNN-based learning methods to model the contextual dependency in the temporal domain. In this paper, we extend this idea to spatio-temporal domains to analyze the hidden sources of action-related information within the input data over both domains concurrently. Inspired by the graphical structure of the human skeleton, we further propose a more powerful tree-structure based traversal method. To handle the noise and occlusion in 3D skeleton data, we introduce new gating mechanism within LSTM to learn the reliability of the sequential input data and accordingly adjust its effect on updating the long-term context information stored in the memory cell. Our method achieves state-of-the-art performance on 4 challenging benchmark datasets for 3D human action analysis.

cs.CV↗

arXiv2016

Test the chiral magnetic effect with isobaric collisions

The quark-gluon matter produced in relativistic heavy-ion collisions may contain local domains in which P and CP symmetries are not preserved. When coupled with an external magnetic field, such P- and CP-odd domains will generate electric currents along the magnetic field --- a phenomenon called the chiral magnetic effect (CME). Recently, the STAR Collaboration at RHIC and the ALICE Collaboration at the LHC released data of charge-dependent azimuthal-angle correlators with features consistent with the CME expectation. However, the experimental observable is contaminated with significant background contributions from elliptic-flow-driven effects, which makes the interpretation of the data ambiguous. In this Letter, we show that the collisions of isobaric nuclei, $^{96}_{44}$Ru + $^{96}_{44}$Ru and $^{96}_{40}$Zr + $^{96}_{40}$Zr, provide an ideal tool to disentangle the CME signal from the background effects. Our simulation demonstrates that the two collision types at $\sqrt{s_{\rm NN}}=200$ GeV have more than $10\%$ difference in the CME signal and less than $2\%$ difference in the elliptic-flow-driven backgrounds for the centrality range of $20-60\%$.

nucl-th↗

arXiv2016

Hierarchical Spatial Sum-Product Networks for Action Recognition in Still Images

Recognizing actions from still images is popularly studied recently. In this paper, we model an action class as a flexible number of spatial configurations of body parts by proposing a new spatial SPN (Sum-Product Networks). First, we discover a set of parts in image collections via unsupervised learning. Then, our new spatial SPN is applied to model the spatial relationship and also the high-order correlations of parts. To learn robust networks, we further develop a hierarchical spatial SPN method, which models pairwise spatial relationship between parts inside sub-images and models the correlation of sub-images via extra layers of SPN. Our method is shown to be effective on two benchmark datasets.

cs.CV↗