arXiv subjects

Feng Jiang

Publications and source records attributed to Feng Jiang.

At least 127 records · Page 7Linked to original sources

arXiv2020

Deep-VFX: Deep Action Recognition Driven VFX for Short Video

Human motion is a key function to communicate information. In the application, short-form mobile video is so popular all over the world such as Tik Tok. The users would like to add more VFX so as to pursue creativity and personlity. Many special effects are added on the short video platform. These gives the users more possibility to show off these personality. The common and traditional way is to create the template of VFX. However, in order to synthesis the perfect, the users have to tedious attempt to grasp the timing and rhythm of new templates. It is not easy-to-use especially for the mobile app. This paper aims to change the VFX synthesis by motion driven instead of the traditional template matching. We propose the AI method to improve this VFX synthesis. In detail, in order to add the special effect on the human body. The skeleton extraction is essential in this system. We also propose a novel form of LSTM to find out the user's intention by action recognition. The experiment shows that our system enables to generate VFX for short video more easier and efficient.

cs.CV↗

arXiv2020

Improving Entity Linking by Modeling Latent Entity Type Information

Existing state of the art neural entity linking models employ attention-based bag-of-words context model and pre-trained entity embeddings bootstrapped from word embeddings to assess topic level context compatibility. However, the latent entity type information in the immediate context of the mention is neglected, which causes the models often link mentions to incorrect entities with incorrect type. To tackle this problem, we propose to inject latent entity type information into the entity embeddings based on pre-trained BERT. In addition, we integrate a BERT-based entity similarity score into the local context model of a state-of-the-art model to better capture latent entity type information. Our model significantly outperforms the state-of-the-art entity linking models on standard benchmark (AIDA-CoNLL). Detailed experiment analysis demonstrates that our model corrects most of the type errors produced by the direct baseline.

cs.CL↗

arXiv2019

Sensor-Movement-Robust Angle Estimation for 3-DoF Lower Limb Joints Without Calibration

Inertial measurement unit (IMU)-based 3-DoF angle estimation methods for lower limb joints have been studied for decades, however the calibration motions and/or careful sensor placement are still necessary due to challenges of real-time application. This study proposes a novel sensormovement-robust 3-DoF method for lower-limb joint angle estimation without calibration. A realtime optimization process, which is based on a feedback iteration progress to identify three joint axes of a 3-DoF joint, has been presented with a reference frame calibration algorithm, and a safe-guarded strategy is proposed to detect and compensate for the errors caused by sensor movements. The experimental results obtained from a 3-DoF gimbal and ten healthy subjects demonstrate a promising performance on 3-DoF angle estimation. Specially, the experiments on ten subjects are performed with three gait modes and a 2-min level walking. The root mean square error is below 2 deg for level walking and 5 deg for other two gait modes. The result of 2-min level walking shows our algorithms stability under long run. The robustness against sensor movement are demonstrated through data from multiple sets of IMUs. In addition, results from the 3-DoF gimbal indicate that the accuracy of 3-DoF angle estimation could be improved by 84.9% with our reference frame calibration algorithm. In conclusion, our study proposes and validates a sensor-movement-robust 3-DoF angle estimation for lowerlimb joints based on IMU. To the best of our knowledge, our approach is the first experimental implementation of IMUbased 3-DoF angle estimation for lower-limb joints without calibration.

eess.SP↗

arXiv2019

Optimized Transmission for Parameter Estimation in Wireless Sensor Networks

A central problem in analog wireless sensor networks is to design the gain or phase-shifts of the sensor nodes (i.e. the relaying configuration) in order to achieve an accurate estimation of some parameter of interest at a fusion center, or more generally, at each node by employing a distributed parameter estimation scheme. In this paper, by using an over-parametrization of the original design problem, we devise a cyclic optimization approach that can handle tuning both gains and phase-shifts of the sensor nodes, even in intricate scenarios involving sensor selection or discrete phase-shifts. Each iteration of the proposed design framework consists of a combination of the Gram-Schmidt process and power method-like iterations, and as a result, enjoys a low computational cost. Along with formulating the design problem for a fusion center, we further present a consensus-based framework for decentralized estimation of deterministic parameters in a distributed network, which results in a similar sensor gain design problem. The numerical results confirm the computational advantage of the suggested approach in comparison with the state-of-the-art methods---an advantage that becomes more pronounced when the sensor network grows large.

eess.SP↗

arXiv2019

Raman spectroscopy of sodium chloride under high-pressure and high-temperature

The high-pressure and high-temperature behaviors of sodium chloride (NaCl) have the fundamental and application significance to the high pressure physics and chemistry. To explore the reactivity of NaCl at high pressure and high temperature, we compressed the NaCl samples to pressures up to 85 GPa and heated them to a temperature of 1800 by using the diamond anvil cell and the double-sided laser heating technology. Our investigation of Raman spectroscopy shows that NaCl decomposes to NaCl3 with orthorhombic Pnma structure plus possible Na3Cl with tetragonal P4/mmm structure above 31 GPa. It shows that the competition of the unusual Cl-Cl covalent bond and Na-Na metallic bond with the Na-Cl ionic bond at high pressure and high temperature may prefer the occurrence of the NaCl3 and Na3Cl combination. We speculate that the large temperature gradient in the NaCl sample layers in the diamond anvil cell experiments, which indicate a non-equilibrium state, is critical to the decomposition reaction of NaCl observed in this study. The catalytic effect of the used laser absorbing materials of transition metal oxides could be another factor to promote the chemical reaction of NaCl. The results of this investigation provide a new pathway to investigate the distinctive non-1:1 stoichiometric alkali metal-halogen compounds that may have exotic properties.

cond-mat.mtrl-sci↗

arXiv2019

Raman spectra of Nontraditional Compound KO4 and Novel Chemical Reaction of KCl-O2

The non-traditional compound KO4 still existing at ambient conditions was synthesized by the reaction of KCl-O2 under high pressure and high temperature (HPHT) using the diamond anvil cell and the laser heating technology. The KCl-O2 sample was heated at 37 GPa (1800 K) and then the products were measured by Raman technology at room temperature. The acquired 1386 cm-1 characteristic vibration band of KO4 reflects the O-O pair with fractional negative charge. The other reaction products include another non-traditional compound KCl3, a bit KClO4 and intermediate product Cl2. The KCl3 can distinguish 11 Raman peaks and decompose into KCl and Cl2 below 10 GPa during decompression. These novel products show that high pressure promotes oxygen and chlorine forming unconventional pair-anions and polyanions, the novel chemical reaction provides new perspectives for developing new batteries materials and studying batteries reactions.

cond-mat.mtrl-sci↗

arXiv2018

Convolutional Neural Networks based Intra Prediction for HEVC

Traditional intra prediction methods for HEVC rely on using the nearest reference lines for predicting a block, which ignore much richer context between the current block and its neighboring blocks and therefore cause inaccurate prediction especially when weak spatial correlation exists between the current block and the reference lines. To overcome this problem, in this paper, an intra prediction convolutional neural network (IPCNN) is proposed for intra prediction, which exploits the rich context of the current block and therefore is capable of improving the accuracy of predicting the current block. Meanwhile, the predictions of the three nearest blocks can also be refined. To the best of our knowledge, this is the first paper that directly applies CNNs to intra prediction for HEVC. Experimental results validate the effectiveness of applying CNNs to intra prediction and achieved significant performance improvement compared to traditional intra prediction methods.

cs.CV↗

arXiv2018

Deep neural network based sparse measurement matrix for image compressed sensing

Gaussian random matrix (GRM) has been widely used to generate linear measurements in compressed sensing (CS) of natural images. However, there actually exist two disadvantages with GRM in practice. One is that GRM has large memory requirement and high computational complexity, which restrict the applications of CS. Another is that the CS measurements randomly obtained by GRM cannot provide sufficient reconstruction performances. In this paper, a Deep neural network based Sparse Measurement Matrix (DSMM) is learned by the proposed convolutional network to reduce the sampling computational complexity and improve the CS reconstruction performance. Two sub networks are included in the proposed network, which are the sampling sub-network and the reconstruction sub-network. In the sampling sub-network, the sparsity and the normalization are both considered by the limitation of the storage and the computational complexity. In order to improve the CS reconstruction performance, a reconstruction sub-network are introduced to help enhance the sampling sub-network. So by the offline iterative training of the proposed end-to-end network, the DSMM is generated for accurate measurement and excellent reconstruction. Experimental results demonstrate that the proposed DSMM outperforms GRM greatly on representative CS reconstruction methods

cs.CV↗

arXiv2018

An efficient deep convolutional laplacian pyramid architecture for CS reconstruction at low sampling ratios

The compressed sensing (CS) has been successfully applied to image compression in the past few years as most image signals are sparse in a certain domain. Several CS reconstruction models have been proposed and obtained superior performance. However, these methods suffer from blocking artifacts or ringing effects at low sampling ratios in most cases. To address this problem, we propose a deep convolutional Laplacian Pyramid Compressed Sensing Network (LapCSNet) for CS, which consists of a sampling sub-network and a reconstruction sub-network. In the sampling sub-network, we utilize a convolutional layer to mimic the sampling operator. In contrast to the fixed sampling matrices used in traditional CS methods, the filters used in our convolutional layer are jointly optimized with the reconstruction sub-network. In the reconstruction sub-network, two branches are designed to reconstruct multi-scale residual images and muti-scale target images progressively using a Laplacian pyramid architecture. The proposed LapCSNet not only integrates multi-scale information to achieve better performance but also reduces computational cost dramatically. Experimental results on benchmark datasets demonstrate that the proposed method is capable of reconstructing more details and sharper edges against the state-of-the-arts methods.

cs.CV↗

arXiv2017

An End-to-End Compression Framework Based on Convolutional Neural Networks

Deep learning, e.g., convolutional neural networks (CNNs), has achieved great success in image processing and computer vision especially in high level vision applications such as recognition and understanding. However, it is rarely used to solve low-level vision problems such as image compression studied in this paper. Here, we move forward a step and propose a novel compression framework based on CNNs. To achieve high-quality image compression at low bit rates, two CNNs are seamlessly integrated into an end-to-end compression framework. The first CNN, named compact convolutional neural network (ComCNN), learns an optimal compact representation from an input image, which preserves the structural information and is then encoded using an image codec (e.g., JPEG, JPEG2000 or BPG). The second CNN, named reconstruction convolutional neural network (RecCNN), is used to reconstruct the decoded image with high-quality in the decoding end. To make two CNNs effectively collaborate, we develop a unified end-to-end learning algorithm to simultaneously learn ComCNN and RecCNN, which facilitates the accurate reconstruction of the decoded image using RecCNN. Such a design also makes the proposed compression framework compatible with existing image coding standards. Experimental results validate that the proposed compression framework greatly outperforms several compression frameworks that use existing image coding standards with state-of-the-art deblocking or denoising post-processing methods.

cs.CV↗

arXiv2017

Deep Networks for Compressed Image Sensing

The compressed sensing (CS) theory has been successfully applied to image compression in the past few years as most image signals are sparse in a certain domain. Several CS reconstruction models have been recently proposed and obtained superior performance. However, there still exist two important challenges within the CS theory. The first one is how to design a sampling mechanism to achieve an optimal sampling efficiency, and the second one is how to perform the reconstruction to get the highest quality to achieve an optimal signal recovery. In this paper, we try to deal with these two problems with a deep network. First of all, we train a sampling matrix via the network training instead of using a traditional manually designed one, which is much appropriate for our deep network based reconstruct process. Then, we propose a deep network to recover the image, which imitates traditional compressed sensing reconstruction processes. Experimental results demonstrate that our deep networks based CS reconstruction method offers a very significant quality improvement compared against state of the art ones.

cs.CV↗

arXiv2017

Single Image Super-Resolution with Dilated Convolution based Multi-Scale Information Learning Inception Module

Traditional works have shown that patches in a natural image tend to redundantly recur many times inside the image, both within the same scale, as well as across different scales. Make full use of these multi-scale information can improve the image restoration performance. However, the current proposed deep learning based restoration methods do not take the multi-scale information into account. In this paper, we propose a dilated convolution based inception module to learn multi-scale information and design a deep network for single image super-resolution. Different dilated convolution learns different scale feature, then the inception module concatenates all these features to fuse multi-scale information. In order to increase the reception field of our network to catch more contextual information, we cascade multiple inception modules to constitute a deep network to conduct single image super-resolution. With the novel dilated convolution based inception module, the proposed end-to-end single image super-resolution network can take advantage of multi-scale information to improve image super-resolution performance. Experimental results show that our proposed method outperforms many state-of-the-art single image super-resolution methods.

cs.CV↗

arXiv2016

Capture the growth kinetics of CVD growth of two-dimensional MoS2

Understanding the microscopic mechanism of chemical vapor deposition (CVD) growth of two-dimensional molybdenum disulfide (2D MoS2) is a fundamental issue towards the function-oriented controlled growth. In this work, we report results on revealing the growth kinetics of 2D MoS2 via capturing the nucleation seed, evolution morphology, edge structure and terminations at the atomic scale during CVD growth using the transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM) studies. The direct growth of few- and mono-layer MoS2 onto graphene based TEM grids allow us to perform the subsequent TEM characterization without any solution-based transfer. Two forms of seeding centers are observed during characterizations: (i) Mo-oxysulfide (MoOxS2-y) nanoparticles either in multi-shelled fullerene-like structures or in compact nanocrystals for the growth of fewer-layer MoS2; (ii) Mo-S atomic clusters in case of monolayer MoS2. In particular, for the monolayer case, at the early stage growth, the morphology appears in irregular polygon shape comprised with two primary edge terminations: S-Mo Klein edge and Mo zigzag edge, approximately in equal numbers, while as the growth proceeds, the morphology further evolves into near-triangle shape in which Mo zigzag edge predominates. Results from density-functional theory calculations are also consistent with the inferred growth kinetics, and thus supportive to the growth mechanism we proposed. In general, the growth mechanisms found here should also be applicable in other 2D materials, such as MoSe2, WS2 and WSe2 etc.

cond-mat.mtrl-sci↗

arXiv2015

Numerical methods for spin-dependent transport calculations and spin bound states analysis in Rashba waveguides

Numerical methods are developed in the quantum transport calculations for electron in the waveguides with spin-orbital (Rashba) interaction. The methods are based on a hybrid mode-matching scheme in which the wavefunctions are expressed as the superposition of eigenmodes in the lead regions and in the device region the wavefunction is expressed on the discrete basis. Two versions are presented for the lead without and with the Rashba interaction. In the latter case the eigenmodes are obtained from a quadratic eigenproblem calculation. These methods are suitable for the systems with variable geometries or arbitrary potential profiles. The computation can be effectively accelerated by the sparse matrix technique. We also investigate the Fano-Rashba bound states in the Rashba waveguides by some nonlinear eigenstate calculation. This calculation is based on a mode-matching method and self-consistent results are obtained in our calculations.

physics.comp-ph↗

arXiv2015

Massive MIMO for Wireless Sensing with a Coherent Multiple Access Channel

We consider the detection and estimation of a zero-mean Gaussian signal in a wireless sensor network with a coherent multiple access channel, when the fusion center (FC) is configured with a large number of antennas and the wireless channels between the sensor nodes and FC experience Rayleigh fading. For the detection problem, we study the Neyman-Pearson (NP) Detector and Energy Detector (ED), and find optimal values for the sensor transmission gains. For the NP detector which requires channel state information (CSI), we show that detection performance remains asymptotically constant with the number of FC antennas if the sensor transmit power decreases proportionally with the increase in the number of antennas. Performance bounds show that the benefit of multiple antennas at the FC disappears as the transmit power grows. The results of the NP detector are also generalized to the linear minimum mean squared error estimator. For the ED which does not require CSI, we derive optimal gains that maximize the deflection coefficient of the detector, and we show that a constant deflection can be asymptotically achieved if the sensor transmit power scales as the inverse square root of the number of FC antennas. Unlike the NP detector, for high sensor power the multi-antenna ED is observed to empirically have significantly better performance than the single-antenna implementation. A number of simulation results are included to validate the analysis.

cs.IT↗

arXiv2015

Efficient Divide-And-Conquer Classification Based on Feature-Space Decomposition

This study presents a divide-and-conquer (DC) approach based on feature space decomposition for classification. When large-scale datasets are present, typical approaches usually employed truncated kernel methods on the feature space or DC approaches on the sample space. However, this did not guarantee separability between classes, owing to overfitting. To overcome such problems, this work proposes a novel DC approach on feature spaces consisting of three steps. Firstly, we divide the feature space into several subspaces using the decomposition method proposed in this paper. Subsequently, these feature subspaces are sent into individual local classifiers for training. Finally, the outcomes of local classifiers are fused together to generate the final classification results. Experiments on large-scale datasets are carried out for performance evaluation. The results show that the error rates of the proposed DC method decreased comparing with the state-of-the-art fast SVM solvers, e.g., reducing error rates by 10.53% and 7.53% on RCV1 and covtype datasets respectively.

cs.LG↗

arXiv2014

Complex absorbing potential based Lorentzian fitting scheme and time dependent quantum transport

Based on the complex absorbing potential (CAP) method, a Lorentzian expansion scheme is developed to express the self-energy. The CAP-based Lorentzian expansion of self-energy is employed to solve efficiently the Liouville-von Neumann equation of one-electron density matrix. The resulting method is applicable for both tight-binding and first-principles models, and is used to simulate the transient currents through graphene nanoribbons and a benzene molecule sandwiched between two carbon-atom-chains.

cond-mat.mes-hall↗

arXiv2014

Energy spread and current-current correlation in quantum systems

We consider energy (heat) transport in quantum systems, and establish a relationship between energy spread and energy current-current correlation function. The energy current-current correlation is related to thermal conductivity by the Green-Kubo formula, and thus this relationship allows us to study conductivity directly from the energy spread process. As an example, we investigate a spinless fermion model; the numerical results confirm the relationship.

cond-mat.stat-mech↗