Search

In this paper, we propose to use the data augmentation of batch drop-block with varying dropping ratios for constructing diversity-achieving branches in person re-identification. Since a considerable portion of input images may be dropped, this reinforces feature learning of the un-dropped region but makes the training process hard to converge. Hence, we propose a novel double-batch-split co-training approach for remedying this problem. In particular, we show that the feature diversity can be well achieved with the use of multiple dropping branches by setting individual dropping ratio for each branch. Empirical evidence demonstrates that the proposed method performs competitively on popular person Re-ID datasets, including Market-1501, DukeMTMC-reID and CUHK03, and the use of more dropping branches can further boost the performance. Source code is available at url.

Deep learning models are widely used in the field of Device-Free Localization (DFL), and their performances rely heavily on the labeled dataset. In this letter, a novel data preprocessing method is proposed for exploiting the full potential of a collected dataset in DFL. Firstly, we propose to employ a pix-level multimodal representation of radio images for a fusion of both amplitude and phase features from Channel State Information (CSI). Then, a novel Conditional Generative Adversarial Network with Auxiliary Classifier (AC-GAN) model is used to generate artificial samples for further expanding the collected dataset. We further employ a regression formulation to train the Convolutional Neural Networks (CNNs) for positioning. The experimental results show that using the proposed method gains approximately 18% Root-Mean-Square-Error (RMSE) improvement over the CSI amplitude-only method and 22% over the CSI phase-only method.

In this paper, a novel decoupled device-free activity detection and position estimation scheme is proposed using convolutional neural networks and channel state information (CSI) measurement as inputs. For the proposed scheme, the two processes of activity recognition and localization are realized in parallel but independently. Compared with the existing joint approaches, the proposed decoupled scheme is free of error propagation between two processes and achieves better performance especially for position estimation. We also propose a CSI based radio image construction using the amplitude measurement with temporal, spatial and frequency domain information. This has been proven very competitive for feature extraction, compared with the state-of-the-art methods. Extensive experimental and simulation results under a real test setup show the superiority of the proposed scheme.

Device activity detection has been extensively investigated for grant-free massive machine-type communications. Instead of using deep Multi-Layer Perception (MLP) networks, this letter proposes a novel convolutional neural network (CNN) architecture for learning device activity from multiple-measurement vectors (MMV). With the use of $1\times 1$ convolutional layers, the proposed CNN could exploit the full potential of shared sparsity among multiple measurements. Extensive simulations show that the proposed CNN outperforms the existing deep MLP network in both performance and computational complexity, especially when the number of measurements increases.

The most recently-proposed concept of capsule network consists of capsules, a structural group of neurons with activation, as building blocks, and dynamic routing between them as connections. Semantic information for final tasks can be extracted by stacking such capsule layers as construction of deep models. In this paper, we formulate the dynamic routing problem from the perspective of feature compression and reconstruction, proposing a novel routing algorithm which explicitly minimizes the reconstruction error with sparsity constraints. By alternately updating poses and routing weights with closed-form solutions, the method ensures a theoretical convergence of the dynamic routing. Benefitting from the strict convergence, we further prompt the efficiency of routing with the use of an adaptive number of iterations. Experimental results on different datasets prove that our proposed routing method outperforms existing routing methods on both performance as well as efficiency with obvious margin. Furthermore, we empirically show that the reconstruction based routing method can be embedded into backbone networks as a representative compression module. The integration of convolutional networks and capsule routing delivers both reduction of model parameters and improvement of performance to original backbones.

The thieno[3,2-b]thiophene (TT) unit is a classic building block to construct high-performance fused-ring electron acceptors (FREAs) for organic solar cells (OSCs). In this work, a star-shaped FREA BTTCTT-ICF based on a new C3h-symmetric fused-ring core BTTCTT was designed and synthesized by introducing the TT unit to extend the π-conjugation. Due to its strong electron-donating ability and efficient π-electron delocalization, BTTCTT-ICF exhibited a narrow optical bandgap of 1.54 eV and a high absorption extinction coefficient of 3.6 × 105 M−1 cm−1. The blend film of BTTCTT-ICF and the wide bandgap polymer PM6 produced a complementary absorption and a broad EQE spectral range covering 300 to 850 nm. When treated via thermal annealing, the PM6:BTTCTT-ICF blend film exhibited more ordered molecular packing and a suitable phase morphology, which improved its charge mobility and device efficiency. Consequently, the PM6:BTTCTT-ICF-based inverted device achieved the best power conversion efficiency (PCE) of 10.39% with a high short-circuit current density (Jsc) of 16.77 mA cm−2, open-circuit voltage (Voc) of 0.91 V and fill factor (FF) of 68.08%. Our results demonstrate that the extension of the star-shaped fused-ring core by TT units is a rational and effective strategy for achieving highly efficient star-shaped FREAs.

Learning part-based features with multiple branches has been proven as an effective way to deliver high performance person re-identification. Existing works mostly exploit extra constraints on different branches to ensure the diversity of extracted features, which may lead to the increased complexity in network architecture and the difficulty for training. In this letter, we propose a quite simple multi-branch structure consisting of a global branch as well as a part branch in an asymmetrical way. We empirically demonstrate that such simple architecture can provide surprisingly high performance without imposing any extra constraint. On top of this, we further prompt the performance with a lightweight implementation of attention module. Extensive experimental results prove that the proposed method, termed Asymmetrical Network (AsNet), outperforms state-of-the-art methods with obvious margin on standard benchmark datasets such as Market1501, DukeMTMC, CUHK03. We believe that AsNet can serve as a strong baseline for related research and the source code is publicly available at https://github.com/www0wwwjs1/asnet.git .

For WiFi indoor localization, collecting the data to construct the fingerprint dataset is a time-consuming and laborious job. In this paper, we propose to alleviate the time-intensive data-labeling by generating artificial fingerprint using a single AC-GAN. Compared to the existing multi-GAN approaches, we show that the use of a single AC-GAN could achieve the improved accuracy in localization, which is validated by extensive experiments. We also discuss its potential limit.

For image semantic segmentation, a fully convolutional network is usually employed as the encoder to abstract visual features of the input image. A meticulously designed decoder is used to decoding the final feature map of the backbone. The output resolution of backbones which are designed for image classification task is too low to match segmentation task. Most existing methods for obtaining the final high-resolution feature map can not fully utilize the information of different layers of the backbone. To adequately extract the information of a single layer, the multi-scale context information of different layers, and the global information of backbone, we present a new attention-augmented module named Dense-attention Context Module (DCM), which is used to connect the common backbones and the other decoding heads. The experiments show the promising results of our method on Cityscapes dataset.

This paper introduces a lightweight convolutional neural network, called FDDWNet, for real-time accurate semantic segmentation. In contrast to recent advances of lightweight networks that prefer to utilize shallow structure, FDDWNet makes an effort to design more deeper network architecture, while maintains faster inference speed and higher segmentation accuracy. Our network uses factorized dilated depth-wise separable convolutions (FDDWC) to learn feature representations from different scale receptive fields with fewer model parameters. Additionally, FDDWNet has multiple branches of skipped connections to gather context cues from intermediate convolution layers. The experiments show that FDDWNet only has 0.8M model size, while achieves 60 FPS running speed on a single RTX 2080Ti GPU with a 1024 × 512 input image. The comprehensive experiments demonstrate that our model achieves state-of-the-art results in terms of available speed and accuracy trade-off on CityScapes and CamVid datasets.

Recently, exploring features from different layers in fully convolutional networks (FCNs) has gained substantial attention to capture context information for semantic segmentation. This paper presents a novel encoder-decoder architecture, called contextual ensemble network (CENet), for semantic segmentation, where the contextual cues are aggregated via densely usampling the convolutional features of deep layer to the shallow deconvolutional layers. The proposed CENet is trained in terms of end-to-end segmentation to match the resolution of input image, and allows us to fully explore contextual features through ensemble of dense deconvolutions. We evaluate our CENet on two widely-used semantic segmentation datasets: PASCAL VOC 2012 and CityScapes. The experimental results demonstrate our CENet achieves superior performance with respect to recent state-of-the-art results. Furthermore, we also evaluate CENet on MS COCO dataset and ISBI 2012 dataset for the task of instance segmentation and biological segmentation, respectively. The experimental results show that CENet obtains promising results on these two datasets.

Visual saliency detection plays a significant role in the fields of computer vision. In this paper, we introduce a novel saliency detection method based on weighted linear multiple kernel learning (WLMKL) framework, which is able to adaptively combine different contrast measurements in a supervised manner. As most influential factor is contrast operation in bottom-up visual saliency, an average weighted corner-surround contrast (AWCSC) is first designed to measure local visual saliency. Combined with common-used center-surrounding contrast (CESC) and global contrast (GC), three types of contrast operations are fed into our WLMKL framework to produce the final saliency map. We show that the assigned weights for each contrast feature maps are always normalized in our WLMKL formulation. In addition, the proposed approach benefits from the advantages of the contribution of each individual contrast feature maps, yielding more robust and accurate saliency maps. We evaluated our method for two main visual saliency detection tasks: human fixed eye prediction and salient object detection. The extensive experimental results show the effectiveness of the proposed model, and demonstrate the integration is superior than individual subcomponent.

Unsupervised domain adaptation (DA) enables intelligent models to learn transferable knowledge from a labeled source domain and adapt to a similar but unlabeled target domain. Studies showed that knowledge could be transferred from one source domain to another unknown target domain, called Universal DA (UDA). However, there is often more than one source domain in the real-world application to be exploited for DA. In this paper, we formally propose a more general domain adaptation setting for image classification, universal multi-source DA (UMDA), where the label sets of multiple source domains can be different, and the label set of the target domain is completely unknown. The main challenge in UMDA is to identify the common label set among each source and target domain and keep the model scalable as the number of source domains increases. In the face of this challenge, we propose a universal multi-source adaptation network (UMAN) to solve the DA problem without increasing the complexity of the model in various UMDA settings. In UMAN, the reliability of each known class belonging to the common label set is estimated via a novel pseudo-margin vector and its weighted form, which helps adversarial training better align the distributions of multiple source domains and target domain. Moreover, the theoretical guarantee for UMAN is also provided. Massive experimental results show that existing UDA and multi-source DA (MDA) methods cannot be directly deployed to UMDA, and the proposed UMAN achieves the state-of-the-art performance in various UMDA settings.

Two solution processable triazatruxene-based small molecules (TATC-BP and TATP-BP) with thermally activated delayed fluorescence (TADF), aggregation-induced emission (AIE), and mechanochromic luminescence (MCL) properties are synthesized and characterized. Compared with phenyl-substituted TATP-BP, TATC-BP with hexyl substituents showed a clearer MCL with a larger emission color change from 483 to 542 nm upon mechanical grinding. On the other hand, the solution processed nondoped OLED of TATP-BP showed a better EL performance with a maximal current efficiency of 18.9 cd A−1, a maximum external quantum efficiency (EQE) of 6.0%, and a remarkable small roll-off of 3.3% in EQE even at a luminance of 1000 cd m−2.

As a more practical setting for unsupervised domain adaptation, Universal Domain Adaptation (UDA) is recently introduced, where the structure of the target label set is assumed to be unavailable. One of the significant challenges in UDA is identifying the common label set shared by source and target domains without labels in the target domain. Most previous works identify the common label set according to the intra-domain information. This article proposes a novel inter-domain approach to identify the common label set efficiently, where the pseudo-margins of target samples are defined to estimate each source class’s probability belonging to the common label set. Specifically, a pseudo-margin vector (PMV) is proposed to describe each source class’s reliability belonging to the common label set, equipped with a target pseudo-margin register (TPMR), which stores and updates the PMV during training. We propose an improved universal adaptation network (I-UAN) to perform domain alignment on the commonly-labeled samples by incorporating the PMV. In I-UAN, a class-wise weighting mechanism based on TPMR for source classes, combined with the sample-wise weighting for target samples, efficiently improves the UDA’s performance. Our approach is robust to outliers as it applies a class-wise weighting mechanism in the source domain. Moreover, we provide experimental analysis on the adaptability, universality, and robustness of I-UAN. Extensive experimental results demonstrate that I-UAN works well in various UDA settings and outperforms the state-of-the-art methods by large margins.

It is known that affinity propagation can perform exemplar-based unsupervised image clustering by taking as input similarities between pairs of images and producing a set of exemplars that best represent the images, and then assigning each nonexemplar image to its most appropriate exemplar. However, the clustering performance of affinity propagation is largely limited by the adopted similarity between any pair of images. As the scale invariant feature transform (SIFT) has been widely employed to extract image features, the nonmetric similarity between any pair of images was proposed by “hard” matching of SIFT features (e.g., counting the number of matching SIFT features). In this letter, we notice, however, that the decision of hard matching of SIFT features is binary, which is not necessary for deriving similarities. Hence, we propose a novel measure of similarities by replacing hard matching with the so-called soft matching. Experimental examples show that significant performance gains can be achieved by the resulting affinity propagation algorithm.

In this study, the author investigates the practical limitation of the recently proposed embedded cryptographic signature authentication scheme at the physical layer. By employing the log-likelihood ratio of a tag bit and its approximation, the author shows that the equivalent authentication channel observed by the secondary receiver can be viewed as a binary-input additive white Gaussian noise channel. Then, the sphere-packing lower bound can be employed to show the transmission capability for practical finite-length authentication tags. To achieve the same effective coverage area for both the primary and secondary receivers, it essentially requires efficient low-rate channel coding schemes with near sphere-packing-bound performance at the secondary receiver, which contrasts sharply with the pessimistic conclusion of Jiang et al.

Learning diverse features is key to the success of person re-identification. Various part-based methods have been extensively proposed for learning local representations, which, however, are still inferior to the best-performing methods for person re-identification. This paper proposes to construct a strong lightweight network architecture, termed PLR-OSNet, based on the idea of Part-Level feature Resolution over the Omni-Scale Network (OSNet) for achieving feature diversity. The proposed PLR-OSNet has two branches, one branch for global feature representation and the other branch for local feature representation. The local branch employs a uniform partition strategy for part-level feature resolution but produces only a single identity-prediction loss, which is in sharp contrast to the existing part-based methods. Empirical evidence demonstrates that the proposed PLR-OSNet achieves state-of-the-art performance on popular person Re-ID datasets, including Market1501, DukeMTMC-reID and CUHK03, despite its small model size.

Unsupervised domain adaptation has achieved great progress in the past few years. Nevertheless, most existing methods work in the so-called closed-set scenario, assuming that the classes depicted by the target samples are exactly the same as those of the source domain. In this paper, we tackle the more challenging scenario of open set domain adaptation with a novel end-to-end training approach, where the samples of unknown class can be present in the target domain. Our method employs entropy minimization for performing unsupervised domain adaptation, where unknown samples are aggressively used in training by forcing the classifier to output the probability of 0.5 on the unknown class. Experimental evidence demonstrates that our approach significantly outperforms the state-of-the-art in open set domain adaptation.

Open set domain adaptation by back propagation (OSDA-BP) was recently proposed as a novel end-to-end training approach for tackling the open set scenario, where only a few categories of interest are shared between source and target data. This paper provides an insightful understanding of the binary cross entropy loss employed in OSDA-BP for picking up the potential unknown samples. With this new understanding, we propose to replace the binary cross entropy loss with a symmetrical Kullback Leibler(KL) distance based loss. This improved OSDA-BP method is extensively evaluated over Office-31 dataset and a considerable performance improvement is observed.

This paper presents a novel saliency detection method based on weighted linear multiple kernel learning (WLMKL), which is able to adaptively combine different contrast measurements in a supervised manner. Three commonly used bottom-up visual saliency operations are first introduced, including corner-surround contrast (CSC), center-surround contrast (CESC), and global contrast (GC). Then these contrast measures are fed into our WLMKL framework to produce the final saliency map. We show that the assigned weights for each contrast feature maps are always normalized in our WLMKL formulation. In addition, the proposed approach benefits from the advantages of the contribution of each individual contrast operation, and thus produces more robust and accurate saliency maps. The extensive experimental results show the effectiveness of the proposed model, and demonstrate the combination is superior to individual subcomponent.

Fully convolutional neural networks (FCNs) have shown great success in dense estimation tasks. One key pillar of such progress is mining multi-scale context cues from features in different convolutional layers. This paper introduces contextual aggregating network(CAN), a generic convolutional feature ensembling framework for semantic segmentation. Our framework first captures multi-scale contextual clues by concatenating multi-level feature representation, which carries both coarse semantics and fine details. Then it adaptively integrates stacked features to perform dense pixel estimation. The proposed CAN is trainable end-to-end, and allows us to fully investigate multi-scale context information embedded in images. The experiments show the promising results of our method on PASCAL VOC 2012 and Cityscapes dataset.

Capsules as well as dynamic routing between them are most recently proposed structures for deep neural networks. A capsule groups data into vectors or matrices as poses rather than conventional scalars to represent specific properties of target instance. Besides of pose, a capsule should be attached with a probability (often denoted as activation) for its presence. The dynamic routing helps capsules achieve more generalization capacity with many fewer model parameters. However, the bottleneck that prevents widespread applications of capsule is the expense of computation during routing. To address this problem, we generalize existing routing methods within the framework of weighted kernel density estimation, and propose a fast routing methods. Our method prompts the time efficiency of routing by nearly 40% with negligible performance degradation. By stacking a hybrid of convolutional layers and capsule layers, we construct a network architecture to handle inputs at a resolution of \(64\times {64}\) pixels. The proposed models achieve a parallel performance with other leading methods in multiple benchmarks.

Fully-convolutional neural networks (CNNs) for semantic segmentation dramatically improve performance using end-to-end learning on whole images in a supervised manner. The success of CNNs for semantic segmentation depends heavily on the pixel-level ground truth, which is labor-intensive in general. To partially solve this problem, domain adaptation techniques have been adapted to the two similar tasks for semantic segmentation, one of which is fully-labelled, while the other is unlabelled. Based on the adversarial learning method for domain adaptation in the context of semantic segmentation (AdaptSegNet), this paper proposes to employ the conditional random field (CRF) to refine the output of the segmentation network before domain adaptation. The proposed system fully integrates CRF model with CNNs, making it possible to train the whole system end-to-end with the usual backpropagation algorithm. Extensive experiments demonstrate the effectiveness of our framework under various domain adaptation settings, including synthetic-to-real scenarios.

As a strong electron-donating π-bridge, alkoxy-substituted thiophene has been widely used in low bandgap fused-ring electron acceptors (FREAs), while its application in star-shaped FREAs has never been explored. Herein, a donor-(π-acceptor)3 type low bandgap star-shaped FREA with ethylhexyloxy-thiophene π-bridge, namely TITT-OT-ICF, is synthesized. The enhanced intramolecular charge transfer of TITT-OT-ICF leads to a high-lying HOMO level, and thus a narrower optical band gap of 1.47 eV relative to its counterpart TITT-T-ICF with ethylhexyl-thiophene as π-bridge (1.61 eV). When blending with a wide bandgap polymer donor PBDB-T, EQE spectrum of TITT-OT–ICF–based organic solar cell (OSC) device is extended to 880 nm, while that of PBDB-T: TITT-T–ICF–based device is limited to 800 nm. In spite of similar phase morphologies in both active layers, the broader photoresponse of PBDB-T: TITT-OT–ICF–based device leads to a higher short-circuit current density (JSC) of 16.95 mA cm−2, yielding a higher power conversion efficiency (PCE) of 8.23% in comparison with PBDB-T: TITT-T–ICF–based device (JSC = 14.33 mA cm−2, PCE = 7.53%). These results demonstrate a rational design of low bandgap star-shaped FREAs based on alkoxy-substituted thiophene as the π-bridge for achieving relatively high JSC and PCE.

Person re-identification (Re-ID) aims at matching images of the same pedestrian under different cameras. State-of-the-art unsupervised domain adaptation (UDA) methods for person Re-ID transferred the learned knowledge from the source domain and further finetuned with pseudo-label target domain via clustering. To reduce the influence of noisy pseudo labels, the Mutual Mean-Teaching (MMT) framework was proposed by conducting pseudo label refinery to better model inter-sample relations in the target domain. Currently, MMT requires the use of the two same deep models for cooperative correction of noisy labels. In this paper, we extend the MMT framework to use two heterogeneous models for cooperative error correction. Experiments show that the proposed heterogeneous MMT framework can work well with two different deep models and its performance, however, is largely limited by the weaker model.

The extensive computational burden limits the usage of convolutional neural networks (CNNs) in edge devices for image semantic segmentation, which plays a significant role in many real-world applications, such as augmented reality, robotics, and self-driving. To address this problem, this paper presents an attention-guided lightweight network, namely AGLNet, which employs an encoder–decoder architecture for real-time semantic segmentation. Specifically, the encoder adopts a novel residual module to abstract feature representations, where two new operations, channel split and shuffle, are utilized to greatly reduce computation cost while maintaining higher segmentation accuracy. On the other hand, instead of using complicated dilated convolution and artificially designed architecture, two types of attention mechanism are subsequently employed in the decoder to upsample features to match input resolution. Specifically, a factorized attention pyramid module (FAPM) is used to explore hierarchical spatial attention from high-level output, still remaining fewer model parameters. To delineate object shapes and boundaries, a global attention upsample module (GAUM) is adopted as global guidance for high-level features. The comprehensive experiments demonstrate that our approach achieves state-of-the-art results in terms of speed and accuracy on three self-driving datasets: CityScapes, CamVid, and Mapillary Vistas. AGLNet achieves 71.3%, 69.4%, and 30.7% mean IoU on these datasets with only 1.12M model parameters. Our method also achieves 52 FPS, 90 FPS, and 53 FPS inference speed, respectively, using a single GTX 1080Ti GPU. Our code is open-source and available at https://github.com/xiaoyufenfei/Efficient-Segmentation-Networks .

The recent years have witnessed great advances for semantic segmentation using deep convolutional neural networks (DCNNs). However, a large number of convolutional layers and feature channels lead to semantic segmentation as a computationally heavy task, which is disadvantage to the scenario with limited resources. In this paper, we design an efficient symmetric network, called (ESNet), to address this problem. The whole network has nearly symmetric architecture, which is mainly composed of a series of factorized convolution unit (FCU) and its parallel counterparts. On one hand, the FCU adopts a widely-used 1D factorized convolution in residual layers. On the other hand, the parallel version employs a transform-split-transform-merge strategy in the designment of residual module, where the split branch adopts dilated convolutions with different rate to enlarge receptive field. Our model has nearly 1.6M parameters, and is able to be performed over 62 FPS on a single GTX 1080Ti GPU. The experiments demonstrate that our approach achieves state-of-the-art results in terms of speed and accuracy trade-off for real-time semantic segmentation on CityScapes dataset.

A popular formulation of domain adaptation (DA) is to simultaneously minimize the source risk and the cross-domain discrepancy between the source domain \(\mathcal {D}_s\) and target domain \(\mathcal {D}_t\). However, this is believed to be suboptimal since the shared feature, which is indistinguishable by a domain classifier, could be far from optimum for the purpose of classification. In this paper, we propose an iterative DA framework for directly optimizing the classification error, which provides DA solutions to both unsupervised and semi-supervised scenarios. Instead of directly attacking \(\mathcal {D}_s \rightarrow \mathcal {D}_t\), we employ an iterative self-training approach of \(\mathcal {D}_s + \mathcal {D}_t^{l-1} \rightarrow \mathcal {D}_t^l\) for progressively-labelling of \(\mathcal {D}_t\) with the aim of \(\lim _{l\rightarrow \infty }\mathcal {D}_t^l \approx \mathcal {D}_t\). For unsupervised DA, it performs comparable to the state-of-the-art DA methods. In particular, it performs the best among various unsupervised DA methods for the very difficult task MNIST \(\rightarrow \) SVHN. By employing a few labeled samples in the target domain, we show that it can achieve significantly improved performance. For MNIST \(\rightarrow \) SVHN, the use of 60 labeled samples from SVHN is able to improve the accuracy margin about \(+10\%\) over the state-or-the-art unsupervised DA method. For a comparison with semi-supervised learning methods, it achieves the accuracy margin about \(+30\%\) over Mean Teacher with 60 labeled samples in SVHN.

Recently, exploring multiple feature maps from different layers in fully convolutional networks (FCNs) has gained substantial attention to capture context information for semantic segmentation. This paper presents a novel encoder-decoder architecture, called dense deconvolutional network (DDN), for semantic segmentation, where the feature maps of deeper convolutional layers are densely upsampled for the shallow deconvolutional layers. The proposed DDN is trainable end-to-end, and allows us to fully investigate multiple scale context cues embedded in images. The experimental results show that our DDN outperforms previous FCNs and encoder-decoder networks (EDNs) on PASCAL VOC 2012 dataset.

When Alice and Bob can observe a common wireless radio channel, it is possible to extract secrets between them. As the wireless channel coefficients over time are generally correlated, it is a common challenge to perform decorrelation efficiently for various secret key generation schemes. In the literature, it was reported that the Karhunen-Lo\(\grave{\text {e}}\)ve (K-L) transform for decorrelation is very sensitive in implementation and often the channel covariance at the side of Alice should be finely transferred to Bob, which is very expensive. In this paper, we provide some insights on the sensitivity of K-L transform, which can be well exploited to develop an improved version of K-L transform for removing this sensitivity. Since the K-L transform is very expensive in complexity, we further propose a linear prediction self-filtering approach at both sides for decorrelation. The theoretical analysis show that its performance is insensitive to the impairment of the wireless channel coefficients observed at different sides, which is further validated by simulations.

The state-of-the-art training of deep neural networks requires to normalize the activities of the neurons for accelerating the training process. A standard approach is to employ batch normalization (BN), in which the activations are normalized by the mean and standard deviation of the training mini-batch. To be invertible, BN also introduces an adaptive gain and bias which are applied after the normalization but often before the non-linearity. In this paper, we investigate the effects of learnable parameters, gain and bias, on the training of various typical deep neural nets, including ALL-CNNs, Network In Network (NIN), ResNets. Through extensive experiments, we show that there is no big difference in both training convergence and final test accuracy if we remove the BN layer following the final convolutional layer from a convolutional neural network (CNN) for standard classification tasks. We also observed that without adaptively updating learnable parameters for BN layers, it often requires less time for training of very deep neural nets such as ResNet-101.

Capsules as well as dynamic routing between them are most recently proposed structures for deep neural networks. A capsule groups data into vectors or matrices as poses rather than conventional scalars to represent specific properties of target instance. Besides of pose, a capsule should be attached with a probability (often denoted as activation) for its presence. The dynamic routing helps capsules achieve more generalization capacity with many fewer model parameters. However, the bottleneck that prevents widespread applications of capsule is the expense of computation during routing. To address this problem, we generalize existing routing methods within the framework of weighted kernel density estimation, and propose two fast routing methods with different optimization strategies. Our methods prompt the time efficiency of routing by nearly 40\% with negligible performance degradation. By stacking a hybrid of convolutional layers and capsule layers, we construct a network architecture to handle inputs at a resolution of $64\times{64}$ pixels. The proposed models achieve a parallel performance with other leading methods in multiple benchmarks., Comment: 16 pages, 4 figures, submitted to eccv 2018

We propose a novel physical-layer challenge-response authentication scheme for multi-carrier transmission. This new scheme exploits both the reciprocity and randomness of the phase responses over multi-carrier channels, and ensures strong security since the channel-phase response is very sensitive to the distance between the transmitter and receiver. A binary hypothesis testing approach is formulated for authentication, which shows a close connection to the problem of pseudo-noise (PN) code acquisition. The proposed scheme can achieve desirable authentication performance even at the signal-to-noise ratio of 5 dB for Rayleigh fading channels.

For physical-layer authentication, the authentication tags are often sent concurrently with messages without much bandwidth expansion. In this paper, we present a channel coding approach for physical-layer authentication. The generation of authentication tags can be formulated as an encoding process for an ensemble of codes, where the shared key between Alice and Bob is considered as the input and the message is used to specify a code from the ensemble of codes. Then, we show that the security of physical-layer authentication schemes can be analyzed through decoding and physical-layer authentication schemes can achieve some degree of information-theoretic security.

This paper presents a novel architecture of iterative receivers with two layers of iterations for turbo coded multiple-input and multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems, where soft messages are passed not only between the MIMO detector and the turbo decoder, but also between the two component decoders within the turbo decoder. We first derive the factor graph representation of a turbo coded system as a basic building block for developing the iterative receivers. Then, a new soft message passing schedule over the factor graph is proposed, resulting in the proposed dual-turbo receiver architecture (DTRA). In DTRA, the MIMO detector and the turbo decoder work concurrently, and the soft messages for both layers of iterations are updated instantaneously, instead of the block-based exchange of soft messages in the conventional iterative receivers. In so doing, the processing latency can be greatly reduced while low computational complexity can be achieved.

Constructed wetlands (CWs) are an emerging, environmentally friendly engineering system employed in China. They require lower investment and operation costs while providing higher treatment efficiency and more ecosystem services than conventional wastewater treatment methods. Introduced to China in 1987, CW systems used for wastewater treatment have rapidly increased in number, particularly since the late 1990s. This review summarizes the state-of-the-art application of CW systems for water pollution treatment by reviewing the basics of the technology and its historical development and performance efficiency. Current progress, limitations, future concerns, and the challenges of CW technologies are also discussed. Also highlighted is the need for sufficient and appropriate data to assist in the further development of CW systems and the implementation of integrated “bottom-up” and “top-down” approaches by both the public in general and government bodies in particular.

In the past, two main approaches for the purpose of authentication, including information-theoretic authentication codes and complexity-theoretic message authentication codes (MACs), were almost independently developed. In this paper, we propose a new cryptographic primitive, namely, artificial-noise-aided MACs (ANA-MACs), which can be considered as both computationally secure and information-theoretically secure. For ANA-MACs, we introduce artificial noise to interfere with the complexity-theoretic MACs and quantization is further employed to facilitate packet-based transmission. With a channel coding formulation of key recovery in the MACs, the generation of standard authentication tags can be seen as an encoding process for the ensemble of codes, where the shared key between Alice and Bob is considered as the input and the message is used to specify a code from the ensemble of codes. Then, we show that the introduction of artificial noise in ANA-MACs can be well employed to resist the key recovery attack even if the opponent has an unlimited computing power. Finally, a pragmatic approach for the analysis of ANA-MACs is provided, and we show how to balance the three performance metrics, including the completeness error, the false acceptance probability, and the conditional equivocation about the key. The analysis can be well applied to a class of ANA-MACs, where MACs with Rijndael cipher are employed., 12 pages, 4 figures, submitted for possible publication

In order to improve the identification performance under unknown target population conditions, a new greedy matching pursuit algorithm (GMP) based localization algorithm is proposed. First of all, based on the possible target position estimations by traditional GMP algorithm, a redefined threshold is proposed to choose more possible target positions from the remaining grids. So the missing probability can be improved. Afterwards, the least square (LS) method is utilized to remove several outliers of the target position estimations and then the false alarm probability can be reduced. Simulation results illustrate that the proposed algorithm has better target identification ability than traditional GMP approach in unknown target population scenarios.

The inherent randomness of wireless radio channels can be employed to extract secrets, whenever the channel reciprocity holds. Privacy amplification is an ingredient step for removing the redundancy of key, of which, the mount of information leaked to potential eavesdroppers, is often difficult to determine. In this paper, we propose a secret-key generation scheme without privacy amplification, of which the disagreement of key is coordinated through the reconciliation of sample positions as did in the level-crossing scheme. Compared to the level-crossing scheme, the proposed approach accumulates the reliable samples and further employs multi-bit quantization for improving the rate of secret-key generation. The accumulated samples should be decorrelated for generating random keys. Through extensive simulations, we find that the Karhunen-Loeve transform for decorrelation is very sensitive in implementation.

Recently, we have developed a PHYsical layer Phase Challenge-Response Authentication Scheme (PHY-PCRAS) for independent multicarrier transmission. In this paper, we make a further step by proposing a novel artificial-noise-aided PHY-PCRAS (ANA-PHY-PCRAS) for practical orthogonal frequency division multiplexing (OFDM) transmission, where the Tikhonov-distributed artificial noise is introduced to interfere with the phase-modulated key for resisting potential key-recovery attacks whenever a static channel between two legitimate users is unfortunately encountered. Then, we address various practical issues for ANA-PHY-PCRAS with OFDM transmission, including correlation among subchannels, imperfect carrier and timing recoveries. Among them, we show that the effect of sampling offset is very significant and a search procedure in the frequency domain should be incorporated for verification. With practical OFDM transmission, the number of uncorrelated subchannels is often not sufficient. Hence, we employ a time-separated approach for allocating enough subchannels and a modified ANA-PHY-PCRAS is proposed to alleviate the discontinuity of channel phase at far-separated time slots. Finally, the key equivocation is derived for the worst case scenario. We conclude that the enhanced security of ANA-PHY-PCRAS comes from the uncertainty of both the wireless channel and introduced artificial noise, compared to the traditional challenge-response authentication scheme implemented at the upper layer., 33 pages, 13 figures, submitted for possible publication

We propose a low-rate turbo-Hadamard coding approach for narrow-band interference (NBI) suppression. The proposed approach can be viewed as the employment of a coded spread-spectrum signalling with time-varying spreading sequences and the NBI suppression is naturally achieved in iterative decoding. Compared with the traditional code-aided NBI suppression with direct-sequence spread-spectrum (DS-SS) signalling, the spreading sequences for the turbo-Hadamard coding approach are non-constant in time and determined from the columns of a Hadamard matrix subject to the low-rate coding rule. The low-rate turbo-Hadamard coding approach is sharply compared with the code-aided DS-SS approach with convolutional coding. With a spreading sequence of length 32, it is shown that the proposed turbo-Hadamard approach outperforms the code-aided DS-SS approach even when its transmission information rate is about 5 times higher than that of the coded-aided DS-SS approach.

Space sense random access (SSRA) with collision mitigation is proposed for TDD wireless local area networks (WLANs) with multiple antennas. Before transmitting a packet, the user performs space sense to detect the status of the channel. If the channel is idle, it then transmits a common training sequence for channel estimation. Multiple subslots are available for channel estimation and the user randomly selects one subslot to achieve the possible collision-free channel estimation. It is shown that SSRA removes the requirement of orthogonal training sequences while avoiding possible packet collision and further providing multipacket reception capability. Compared to the conventional slotted ALOHA scheme, SSRA can achieve much higher throughput, as verified by extensive simulation results.

For fast acquisition of GPS pseudo-noise (PN) code, folding is often used to speed up the search process. By properly incorporating the folding in the compressive sensing (CS) framework, a new low-complexity GPS signal acquisition scheme is proposed. In the CS reconstruction step, a new method is proposed termed projection elimination (PE) algorithm, which has lower computational complexity and nearly the same performance compared to the orthogonal matching pursuit (OMP) algorithm for very sparse signals. Simulation results are given to demonstrate its effectiveness.

In this paper, we show how to compute the Slepian-Wolf lower bound for the reconciliation of physical layer secret keys between two radio terminals. Two coding approaches, including the syndrome method and the parity method, have been proven to require the same number of bits exchanged between two terminals for the reconciliation in the noiseless environment. We also present a practical coding approach based on LDPC codes, and the gap between the practical approach and the Slepian-Wolf bound is given.

The reciprocal nature of wireless radio channels can be well employed to extract a secret key between two terminals. The maximum size of a key that can be shared is characterized by the mutual information between the observations of two radios. We consider a correlated fading channel and a closed form for the mutual information is derived. It is found that the secrecy capacity is largely limited by the maximum Doppler frequency.

The generation-based network coding has been successfully employed in overlay networks. The key idea is the generation-marking for packets along with the suitable control of buffering and transmission opportunity at nodes in network. Building on recent work by Chou et al., who present the generation-based network coding for multicast networks, we extend this framework to arbitrary network coding problems over networks with delay. It is shown that generation-based technique can transform the nonzero-delay network into the zero-delay network and however, inappropriate transmission opportunity or short of memory at nodes may change the topology of the original network.

Space-time codes increase the transmission rate of a communication system in fading channels significantly. The decoding of space-time codes generally depends on perfect channel estimation. Differential space-time modulation (DSTM) can work, in a noncoherent manner, over continuously fading channels. However, it exhibits an irreducible error floor in time-correlated fading channels. The impact of correlated Rayleigh fading on DSTM is investigated, and a decision-feedback linear prediction receiver for DSTM is presented. Computer simulations show that the linear prediction receiver reduces the error floor substantially.