In the study of compressed sensing (CS), two main challenges are design sampling matrix and development reconstruction method. On one hand, usually used random matrices (e.g., GRM) signal independent, which ignore characteristics signal. other state-of-the-art image CS methods GSR MH) achieve quite good performance, however with much higher computational complexity. To deal challenges, we propose an framework using convolutional neural network (dubbed CSNet) that includes a network,...

Deep learning methods, e.g., convolutional neural networks (CNNs) and Recurrent Neural Networks (RNNs), have achieved great success in image processing natural language especially high level vision applications such as recognition understanding. However, it is rarely used to solve information security problems attack detection studied this paper. Here, we move forward a step propose novel multi-channel intelligent method based on long short term memory recurrent (LSTM-RNNs). To achieve rate,...

Deep learning, e.g., convolutional neural networks (CNNs), has achieved great success in image processing and computer vision especially high-level applications, such as recognition understanding. However, it is rarely used to solve low-level problems compression studied this paper. Here, we move forward a step propose novel framework based on CNNs. To achieve high-quality at low bit rates, two CNNs are seamlessly integrated into an end-to-end framework. The first CNN, named compact network...

The compressed sensing (CS) theory has been successfully applied to image compression in the past few years as most signals are sparse a certain domain. Several CS reconstruction models have recently proposed and obtained superior performance. However, there still exist two important challenges within theory. first one is how design sampling mechanism achieve an optimal efficiency, second perform get highest quality signal recovery. In this paper, we try deal with these problems deep...

Recently, deep learning based image Compressed Sensing (CS) methods have been proposed and demonstrated superior reconstruction quality with low computational complexity. However, the existing CS need to train different models for sampling ratios, which increases complexity of encoder decoder. In this paper, we propose a scalable convolutional neural network (dubbed SCSNet) achieve only one model. Specifically, SCSNet provides both coarse fine granular scalability. For scalability, is...

In this paper, a new image segmentation method is proposed by combining the FCM clustering algorithm with rough set theory. First, attribute value table constructed based on results of under different numbers, and divided into several small regions indistinguishable relationship attributes. Then, weight values each are obtained reduction used as basis to calculate difference between then similarity evaluation region realized through equivalence defined degree. Finally, final relation merge...

Sparse representation-based visual tracking approaches have attracted increasing interests in the community recent years. The main idea is to linearly represent each target candidate using a set of and trivial templates, while imposing sparsity constraint onto representation coefficients. After we obtain coefficients ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm minimization methods, with lowest error, when it reconstructed...

The traditional classification methods for limb motion recognition based on sEMG have been deeply researched and shown promising results. However, information loss during feature extraction reduces the accuracy. To obtain higher accuracy, deep learning method was introduced. In this paper, we propose a parallel multiple-scale convolution architecture. Compared with state-of-art methods, proposed architecture fully considers characteristics of signal. Larger sizes kernel filter than commonly...

The Brain-computer interface (BCI) is used to enhance the human capabilities. hybrid-BCI (hBCI) a novel concept for subtly hybridizing multiple monitoring schemes maximize advantages of each while minimizing drawbacks individual methods. Recently, researchers have started focusing on Electroencephalogram (EEG) and "Functional Near-Infrared Spectroscopy" (fNIRS) based hBCI. main reason due development artificial intelligence (AI) algorithms such as machine learning approaches better process...

Feature extraction has always been a difficult problem in the classification performance of synthetic aperture radar automatic target recognition (SAR-ATR). It is very important to select discriminative features train classifier, which prerequisite. Inspired by great success convolutional neural network (CNN), we address SAR proposing feature method, takes advantage exploiting extracted deep from CNNs on images introduce more powerful and robust representation ability for them. First,...

Traditional works have shown that patches in a natural image tend to redundantly recur many times inside the image, both within same scale, as well across different scales. Make full use of these multi-scale information can improve restoration performance. However, current proposed deep learning based methods do not take into account. In this paper, we propose dilated convolution inception module learn and design network for single super-resolution. Different learns scale feature, then...

This paper introduces an OSA detection method based on Recurrent Neural network. At the first step, RR interval (time from one R wave to next wave) is employed extract signals Apnea- Electrocardiogram (ECG) where all extracted features are then used as input for designed deep model. Then architecture having four recurrent layers and batch normalization trained with detection. Apnea-ECG datasets physionet.org training testing our Experimental results reveal that automatic model provides...

Existing state of the art neural entity linking models employ attention-based bag-of-words context model and pre-trained embeddings bootstrapped from word to assess topic level compatibility. However, latent type information in immediate mention is neglected, which causes often link mentions incorrect entities with type. To tackle this problem, we propose inject into based on BERT. In addition, integrate a BERT-based similarity score local state-of-the-art better capture information. Our...

Summary form only given. Traditional intra prediction methods for HEVC rely on using the nearest reference lines predicting a block, which ignore much richer context between current block and its neighboring blocks therefore cause inaccurate especially when weak spatial correlation exists lines. To overcome this problem, in paper, an intra-prediction convolutional neural network (IPCNN) is proposed prediction, exploits rich of capable improving accuracy block. Meanwhile, reconstruction three...

Deep network-based image Compressed Sensing (CS) has attracted much attention in recent years. However, the existing deep CS schemes either reconstruct target a block-by-block manner that leads to serious block artifacts or train network as black box brings about limited insights of prior knowledge. In this paper, novel framework using non-local neural (NL-CSNet) is proposed, which utilizes self-similarity priors with improve reconstruction quality. proposed NL-CSNet, two subnetworks are...

Compressive Sensing (CS) theory shows that a signal can be decoded from many fewer measurements than suggested by the Nyquist sampling theory, when is sparse in some domain. Most of conventional CS recovery approaches, however, exploited set fixed bases (e.g. DCT, wavelet, contour let and gradient domain) for entirety signal, which are irrespective nonstationarity natural signals cannot achieve high enough degree sparsity, thus resulting poor rate-distortion performance. In this paper, we...