Search

Sparse-array synthesis can considerably reduce the number of sensor elements while optimizing the beam-pattern response performance. The sparsity of an array is related to the degrees of freedom of the array elements. A sparse-array method based on iterative convex optimization and a simulated-annealing expanded array is proposed in this paper. This method transforms the sparse-array problem into a minimum l1 norm problem and obtains the sparse array by solving the convex optimization problem using the primal-dual algorithm. Meanwhile, to improve the degree of freedom, array elements are expanded using stochastic perturbation. According to the simulated-annealing algorithm, the closed array elements are reopened with a specific probability, which is iteratively thinned and expanded. The results show that the proposed method can obtain an extremely sparse array, which is better than that obtained using the existing methods. [ABSTRACT FROM AUTHOR]

The performance of deep neural networks (DNNs) critically relies on high-quality annotations, while training DNNs with noisy labels remains challenging owing to their incredible capacity to memorize the entire training set. In this work, we use two synchronously trained networks to reveal that noisy labels may result in more divergent gradients when updating the parameters. To overcome this, we propose a novel co-training framework named gradient agreement learning (GAL). By dynamically evaluating the gradient agreement coefficient of every pair of parameters from two identical DNNs to determine whether to update them in the training process. GAL can effectively hinder the memorization of noisy labels. Furthermore, we utilize the pseudo labels produced by the two DNNs as the supervision for the training of another network, thereby gaining further improvement by correcting some noisy labels while overcoming the confirmation bias. Extensive experiments on various benchmark datasets demonstrate the superiority of the proposed GAL. [ABSTRACT FROM AUTHOR]

Automatic detection of epileptic seizures is important in epilepsy control and treatment, and specific feature extraction assists in accurate detection. We developed a feature extraction method for seizure detection based on multi-site synchronous changes and an edge detection algorithm. We investigated five chronic temporal lobe epilepsy rats with 8- and 12-channel detection sites in the hippocampus and limbic system. Multi-site synchronous changes were selected as a specific feature and implemented as a seizure detection method. For preprocessing, we used magnitude-squared coherence maps and Canny edge detection algorithm to find the frequency band with the most significant change in synchronization and the important channel pairs. In detection, we used the maximal cross-correlation coefficient as an indicator of synchronization and the correlation coefficient curves' average value and standard deviation as two detection features. The method achieved high performance, with an average 96.60% detection rate, 2.63/h false alarm rate, and 1.25 s detection delay. The experimental results show that synchronization is an appropriate feature for seizure detection. The magnitude-squared coherence map can assist in selecting a specific frequency band and channel pairs to enhance the detection result. We found that individuals have a specific frequency band that reflects the most significant synchronization changes, and our method can individually adjust parameters and has good detection performance. [ABSTRACT FROM AUTHOR]

A small portion of mislabeled data can easily limit the performance of deep neural networks (DNNs) due to their high capacity for memorizing random labels. Thus, robust learning from noisy labels has become a key challenge for deep learning due to inadequate datasets with high-quality annotations. Most existing methods involve training models on clean sets by dividing clean samples from noisy ones, resulting in large amounts of mislabeled data being unused. To address this problem, we propose categorizing training samples into five fine-grained clusters based on the difficulty experienced by DNN models when learning them and label correctness. A novel fine-grained confidence modeling (FGCM) framework is proposed to cluster samples into these five categories; with each cluster, FGCM decides whether to accept the cluster data as they are, accept them with label correction, or accept them as unlabeled data. By applying different strategies to the fine-grained clusters, FGCM can better exploit training data than previous methods. Extensive experiments on widely used benchmarks CIFAR-10, CIFAR-100, clothing1M, and WebVision with different ratios and types of label noise demonstrate the superiority of our FGCM. [ABSTRACT FROM AUTHOR]

Large uniform rectangular arrays (URAs) are widely used in 3-D underwater imaging systems to obtain high-quality images; however, the gain and phase errors of these arrays can distort the results. To restore the image quality, many calibration methods have been proposed, most of which adopt a far-field model. This model requires the calibrator source to be far away, which reduces the signal-to-noise ratio and adversely affects the calibration accuracy. An effective means of alleviating this contradiction is to adopt the near-field signal model. However, the source location estimation in the near-field model typically uses spectral peak searches, which is a significant computational burden, particularly for large URAs. A fast and efficient near-field model self-calibration method for large URAs is proposed to obtain improved calibration results. First, a convenient preprocessing and two extended three-step iterative algorithms are used instead of spectral peak searches to estimate the source location efficiently. The array is then calibrated using a spatial matched filter and maximum a posteriori approach. Numerical results proved that the proposed method almost reaches the Cramér–Rao bound. Compared with existing near-field model calibration methods using spectral peak searches, the computational complexity of the proposed method is reduced by more than three orders of magnitude when the array size reaches 60 $\times $ 60 or more. By virtue of the near-field model, the phase calibration accuracy for large URAs at a close range is more than 10 times that of far-field model calibration methods. [ABSTRACT FROM AUTHOR]

Transferable adversarial attacks against Deep neural networks (DNNs) have received broad attention in recent years. An adversarial example can be crafted by a surrogate model and then attack the unknown target model successfully, which brings a severe threat to DNNs. The exact underlying reasons for the transferability are still not completely understood. Previous work mostly explores the causes from the model perspective, e.g., decision boundary, model architecture, and model capacity. Here, we investigate the transferability from the data distribution perspective and hypothesize that pushing the image away from its original distribution can enhance the adversarial transferability. To be specific, moving the image out of its original distribution makes different models hardly classify the image correctly, which benefits the untargeted attack, and dragging the image into the target distribution misleads the models to classify the image as the target class, which benefits the targeted attack. Towards this end, we propose a novel method that crafts adversarial examples by manipulating the distribution of the image. We conduct comprehensive transferable attacks against multiple DNNs to demonstrate the effectiveness of the proposed method. Our method can significantly improve the transferability of the crafted attacks and achieves state-of-the-art performance in both untargeted and targeted scenarios, surpassing the previous best method by up to 40% in some cases. In summary, our work provides new insight into studying adversarial transferability and provides a strong counterpart for future research on adversarial defense. [ABSTRACT FROM AUTHOR]

Pneumonia remains a threat to human health; the coronavirus disease 2019 (COVID-19) that began at the end of 2019 had a major impact on the world. It is still raging in many countries and has caused great losses to people's lives and property. In this paper, we present a method based on DeepConv-DilatedNet of identifying and localizing pneumonia in chest X-ray (CXR) images. Two-stage detector Faster R-CNN is adopted as the structure of a network. Feature Pyramid Network (FPN) is integrated into the residual neural network of a dilated bottleneck so that the deep features are expanded to preserve the deep feature and position information of the object. In the case of DeepConv-DilatedNet, the deconvolution network is used to restore high-level feature maps into its original size, and the target information is further retained. On the other hand, DeepConv-DilatedNet uses a popular fully convolution architecture with computation shared on the entire image. Then, Soft-NMS is used to screen boxes and ensure sample quality. Also, K-Means++ is used to generate anchor boxes to improve the localization accuracy. The algorithm obtained 39.23% Mean Average Precision (mAP) on the X-ray image dataset from the Radiological Society of North America (RSNA) and got 38.02% Mean Average Precision (mAP) on the ChestX-ray14 dataset, surpassing other detection algorithms. So, in this paper, an improved algorithm that can provide doctors with location information of pneumonia lesions is proposed. [ABSTRACT FROM AUTHOR]

An adaptive beamformer is effective at suppressing interference and noise. However, when the desired signal component is included in the covariance matrix, the beamformer performance becomes seriously degraded. Moreover, while the linear array has been actively researched, few studies have focused on the planar array. In this paper, an adaptive beamformer with more accurate reconstruction of the covariance matrix for a planar array is therefore proposed. The reconstruction is based on the Bayesian compressive sensing (BCS) theory. First, the directions of arrival (DOA) estimation of interferences are conducted. This problem is transformed into that of finding the minimum number of DOAs with a nonzero input because the array output is known. Accordingly, it can be converted into a probabilistic framework using the BCS technique. Then, the interference plus noise covariance matrix is reconstructed by using the DOAs of the interferences and the Capon spatial spectrum estimator. The reconstruction matrix is more accurate than other methods that directly use a data sampling matrix. Further constraints are then added to control the side-lobe level of the beam pattern of the proposed beamformer. Our numerical results confirm the effectiveness of the proposed method in terms of interference suppression, robustness to mismatch errors, and effective side-lobe-level control. [ABSTRACT FROM AUTHOR]

Several dual-domain convolutional neural network-based methods show outstanding performance in reducing image compression artifacts. However, they are unable to handle color images as the compression processes for gray scale and color images are different. Moreover, these methods train a specific model for each compression quality, and they require multiple models to achieve different compression qualities. To address these problems, we proposed an implicit dual-domain convolutional network (IDCN) with a pixel position labeling map and quantization tables as inputs. We proposed an extractor-corrector framework-based dual-domain correction unit (DCU) as the basic component to formulate the IDCN; the implicit dual-domain translation allows the IDCN to handle color images with discrete cosine transform (DCT)-domain priors. A flexible version of IDCN (IDCN-f) was also developed to handle a wide range of compression qualities. Experiments for both objective and subjective evaluations on benchmark datasets show that IDCN is superior to the state-of-the-art methods and IDCN-f exhibits excellent abilities to handle a wide range of compression qualities with a little trade-off against performance; further, it demonstrates great potential for practical applications. [ABSTRACT FROM AUTHOR]

Since the human face preserves the richest information for recognizing individuals, face recognition has been widely investigated and achieved great success in various applications in the past decades. However, face spoofing attacks (e.g., face video replay attack) remain a threat to modern face recognition systems. Though many effective methods have been proposed for anti-spoofing, we find that the performance of many existing methods is degraded by illuminations. It motivates us to develop illumination-invariant methods for anti-spoofing. In this paper, we propose a two-stream convolutional neural network (TSCNN), which works on two complementary spaces: RGB space (original imaging space) and multi-scale retinex (MSR) space (illumination-invariant space). Specifically, the RGB space contains the detailed facial textures, yet it is sensitive to illumination; MSR is invariant to illumination, yet it contains less detailed facial information. In addition, the MSR images can effectively capture the high-frequency information, which is discriminative for face spoofing detection. Images from two spaces are fed to the TSCNN to learn the discriminative features for anti-spoofing. To effectively fuse the features from two sources (RGB and MSR), we propose an attention-based fusion method, which can effectively capture the complementarity of two features. We evaluate the proposed framework on various databases, i.e., CASIA-FASD, REPLAY-ATTACK, and OULU, and achieve very competitive performance. To further verify the generalization capacity of the proposed strategies, we conduct cross-database experiments, and the results show the great effectiveness of our method. [ABSTRACT FROM AUTHOR]

Synthesis of sparse planar arrays can effectively reduce hardware costs and computational complexity in phased array 3-D imaging sonar systems. Traditional stochastic methods, such as simulated annealing, require multiple experiments and parameter adjustments to obtain optimal results. Methods based on compressed sensing (CS) can overcome this defect. However, when applied to large arrays, CS methods require vast computational complexity and may not obtain optimal sparse results because of violating the restricted isometry property. To make CS methods more practical, a distributed convex optimization CS method is proposed here for the sparse planar array synthesis in 3-D imaging sonar systems. This method is based on the CS theory, solving the minimum number of active elements under certain beam pattern constraints using the iterative reweighted l1-norm minimization algorithm. Then, a multistage distributed framework is proposed to decompose the array into multistage subarrays, and the array synthesis is performed sequentially for each stage subarray to reduce computational complexity and obtain higher sparsity rates. Some applications of sparse planar array synthesis are employed to evaluate the efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

Sparse planar array designs significantly reduce the hardware complexity and computational overhead in phased array 3-D imaging sonar systems. Recently, Bayesian compressive sensing (BCS) theory has been applied for synthesizing maximally sparse arrays, in which case, the problem is formulated as a pattern matching technique in a probabilistic fashion. Thus, in this study, a perturbed BCS-based method with a minimum inter-element spacing constraint is proposed. In particular, a modified BCS technique is first employed to synthesize the primary sparse array. Then, the position perturbations of the elements are determined via the first-order Taylor expansion to increase their degrees of freedom (DOFs). The position perturbations allow for continuous element arrangement, which compensates for the imperfection of discrete candidate sample locations. After that, an improved element merging technique combined with convex optimization was adopted to establish constraints on element spacing while mitigating beam pattern degradation. In practice, this is a feasible approach, which merges elements close to each other, leading to a sparser array with a minimum inter-element spacing. Our numerical results confirm the validity of the proposed method in matching accuracy, array sparsity, and constraining the minimum spacing between elements. [ABSTRACT FROM AUTHOR]

Pneumonia is a disease that develops rapidly and seriously threatens the survival and health of human beings. At present, the computer-aided diagnosis (CAD) of pneumonia is mostly based on binary classification algorithms that cannot provide doctors with location information. To solve this problem, this study proposes an end-to-end highly efficient algorithm for the detection of pneumonia based on a convolutional neural network—Pneumonia Yolo (PYolo). This algorithm is an improved version of the Yolov3 algorithm for X-ray image data of the lungs. Dilated convolution and an attention mechanism are used to improve the detection results of pneumonia lesions. In addition, double K-means is used to generate an anchor box to improve the localization accuracy. The algorithm obtained 46.84 mean average precision (mAP) on the X-ray image dataset provided by the Radiological Society of North America (RSNA), surpassing other detection algorithms. Thus, this study proposes an improved algorithm that can provide doctors with location information on lesions for the detection of pneumonia. [ABSTRACT FROM AUTHOR]

For decades, task functional magnetic resonance imaging has been a powerful noninvasive tool to explore the organizational architecture of human brain function. Researchers have developed a variety of brain network analysis methods for task fMRI data, including the general linear model, independent component analysis, and sparse representation methods. However, these shallow models are limited in faithful reconstruction and modeling of the hierarchical and temporal structures of brain networks, as demonstrated in more and more studies. Recently, recurrent neural networks (RNNs) exhibit great ability of modeling hierarchical and temporal dependence features in the machine learning field, which might be suitable for task fMRI data modeling. To explore such possible advantages of RNNs for task fMRI data, we propose a novel framework of a deep recurrent neural network (DRNN) to model the functional brain networks from task fMRI data. Experimental results on the motor task fMRI data of Human Connectome Project 900 subjects release demonstrated that the proposed DRNN can not only faithfully reconstruct functional brain networks, but also identify more meaningful brain networks with multiple time scales which are overlooked by traditional shallow models. In general, this work provides an effective and powerful approach to identifying functional brain networks at multiple time scales from task fMRI data. [ABSTRACT FROM AUTHOR]

In this paper, we present novel sharp attention networks by adaptively sampling feature maps from convolutional neural networks for person re-identification (re-ID) problems. Due to the introduction of sampling-based attention models, the proposed approach can adaptively generate sharper attention-aware feature masks. This greatly differs from the gating-based attention mechanism that relies on soft gating functions to select the relevant features for person re-ID. In contrast, the proposed sampling-based attention mechanism allows us to effectively trim irrelevant features by enforcing the resultant feature masks to focus on the most discriminative features. It can produce sharper attentions that is more assertive in localizing subtle features relevant to re-identifying people across cameras. For this purpose, a differentiable Gumbel-Softmax sampler is employed to approximate the Bernoulli sampling to train the sharp attention networks. Extensive experimental evaluations demonstrate the superiority of this new sharp attention model for person re-ID over other related existing, published state-of-the-art works on three challenging benchmarks, including CUHK03, Market-1501, and DukeMTMC-reID. [ABSTRACT FROM AUTHOR]

The development of a real-time 3-D underwater imaging system with a planar array is restricted by the huge hardware cost and computational burden associated with the large number of elements required. In this paper, a new design of low-complexity 3-D underwater imaging system is proposed, involving two key parts: a special sparse array with low hardware cost and an efficient beamforming with low computational burden. Especially, the simulated annealing algorithm is optimized based on distributed and parallel subarray (DPS) beamforming and results in fewer active elements than existing methods. To improve the computational efficiency especially for sparse arrays, an optimized DPS beamforming is presented with the pruning fast Fourier transform technology. To test the validity of the design, a real system based on the proposed methods was devised and employed in the lake and sea trials. The experiment results demonstrate that the low-complexity 3-D underwater imaging system can provide a satisfactory imaging quality. [ABSTRACT FROM AUTHOR]

Traditional frame interpolation algorithms typically find dense correspondences to synthesize an in-between frame. Finding correspondences is often sensitive to occlusion, disocclusion, and changes in color or luminance. We present a phase-feature-aided multiframe interpolation network that aims to estimate multiple in-between frames in one pass and handle challenging scenarios such as extreme light changes and occlusion. We first model the relation between multiple in-between frames together to enhance the temporal consistency. Two candidate optical flow fields are produced for a given in-between frame, one predicted from our network and the other estimated from those of neighboring frames using a flow fusion map. We also employ an image fusion map to combat occlusion problems in the warping processes, producing two candidate interpolated images that are fed to a shallow network with a residual structure to obtain the final interpolated image. To handle challenging scenarios, we apply a set of Gabor filters to extract phase variations in the feature domain with a multiscale phase subnetwork. Our entire neural network is end-to-end trainable. Our experiments show that this method outperforms the state-of-the-art approaches and achieves marked visual improvement in challenging scenarios. [ABSTRACT FROM AUTHOR]

Two-dimensional arrays are required to generate 3-D acoustic images in both near-field and far-field regions. However, almost all of the sparse methods, which decrease the number of elements, only synthesize the 2-D sparse arrays in far-field conditions. In this paper, a novel method is proposed for optimizing sparse array in both near-field and far-field conditions that improving the sparseness of the optimized array. First, a new energy function is proposed and successfully synthesizes sparse arrays in both near-field and far-field conditions. The simplified form of the new energy function is then presented for cutting down computational load during optimization process. Second, the nongrid sparse array technology, which connotes irregular element positioning, is introduced into the simulated annealing algorithm to extensively reduce the number of active elements. Finally, the optimized method is demonstrated through designing a cross array with 100 + 100 elements. The results show notable improvements on the array characteristics in random working distances over those reported in the literature. [ABSTRACT FROM AUTHOR]

Brain activity is a dynamic combination of different sensory responses and thus brain activity/state is continuously changing over time. However, the brain’s dynamical functional states recognition at fast time-scales in task fMRI data have been rarely explored. In this paper, we propose a novel 5-layer deep sparse recurrent neural network (DSRNN) model to accurately recognize the brain states across the whole scan session. Specifically, the DSRNN model includes an input layer, one fully-connected layer, two recurrent layers, and a softmax output layer. The proposed framework has been tested on seven task fMRI data sets of Human Connectome Project. Extensive experiment results demonstrate that the proposed DSRNN model can accurately identify the brain’s state in different task fMRI data sets and significantly outperforms other auto-correlation methods or non-temporal approaches in the dynamic brain state recognition accuracy. In general, the proposed DSRNN offers a new methodology for basic neuroscience and clinical research. [ABSTRACT FROM AUTHOR]