Sharif Digital Repository

Developing time-invariant solutions for recognition of human action is still an important and open challenge. Three issues make time-invariant solutions so important: different speed of performing the same action by different people, latency in doing the actions and the existence of redundant frames in the recorded video. To overcome these problems, we propose a method based on the so-called memory of the joints to remember only the cumulative positive and negative movement of each joint. Hence, we transform action recognition from time-space to shape-space and the action recognition becomes the problem of shape classification. These shape features contain highly discriminative information... 

We propose a novel cooperative transmission scheme called Convolutional Network-Coded Cooperation (CNCC) for a network including N sources, one M-antenna relay, and one common destination. The source-relay (S-R) channels are assumed to be Nakagami-m fading, while the source-destination (S-D) and the relay-destination (R-D) channels are considered Rayleigh fading. The CNCC scheme exploits the generator matrix of a good (N + M', N,v) systematic convolutional code, with the free distance of dfree designed over GF(2), as the network coding matrix which is run by the network's nodes, such that the systematic symbols are directly transmitted from the sources, and the parity symbols are sent by the... 

In this paper, an advanced integrated pixel-level method based on the deep convolutional neural network (DCNN) approach named DeepLabv3+ is proposed for weak interlayers detection and quantification. Furthermore, a database containing 32,040 images of limestone, dolomite, loess clay, and red clay is established to verify this method. The proposed model is then trained, validated, and tested via feeding multiple weak interlayers. Moreover, robustness and adaptability of the proposed model are evaluated, and the weak interlayers are extracted. Compared with the fully convolutional network (FCN)-based method and traditional image techniques, the proposed model provides higher accuracy in terms... 

Automatic image annotation is a mechanism to assign a list of appropriate tags that describe the visual content of a given image. Most methods only focus on the content of the images and ignore the relationship between the tags in vocabulary. In this work, we propose a new deep learning-based automatic image annotation architecture, which considers label dependencies in a graph convolution neural network structure and extracts tag descriptors to re-weight the output class scores based on their relationships. The proposed architecture has three main parts: feature extraction, graph convolutional network, and annotation. In graph convolutional network, we apply one layer convolution on... 

Accurately determining the nutritional status of plants can prevent many diseases caused by fertilizer disorders. Leaf analysis is one of the most used methods for this purpose. However, in order to get a more accurate result, disorders must be identified before symptoms appear. Therefore, this study aims to identify leaves with excessive nitrogen using one‐dimensional convolutional neural networks (1D‐CNN) on a dataset of spectral data using the Keras library. Seeds of cucumber were planted in several pots and, after growing the plants, they were divided into different classes of control (without excess nitrogen), N30% (excess application of nitrogen fertilizer by 30%), N60% (60% overdose),... 

In recent years, convolutional neural networks and deep neural networks have been used extensively in various fields of analytical chemistry. The use of these models for calibration tasks has been highly effective; however, few reports have been published on their properties and characteristics of analytical figures of merit. Currently, most performance measures for these types of networks only incorporate some function of prediction error. While useful, these measures are incomplete and cannot be used as an objective comparison among different models. In this report, a new method for calculating the sensitivity of any type of neural network is proposed and studied on both simulated and real... 

As a critical part of any security system, identity recognition has become paramount among researchers. In this regard, several methods are presented while considering various sensors and data. In particular, gait data yields rich information about a person, including some exclusive moving patterns which can be utilized to distinguish between different individuals. On the other hand, convolutional neural networks are proved to be applicable for structured data, especially images. In this article, 12 markers are considered in gathering the gait data, each representing a lower-body joint location. Then, utilizing the gait data in a 2D tensor form, three different convolutional neural networks... 

In this paper, we consider a low-complexity iterative receiver lor decoding multi-user information data in a bandwith efficient convolutionally coded TH-UWB-CDMA system introduced in [1]. The proposed receiver structure consists of a bank of soft interference canceller likelihood calculators (SICLC). each followed by a soft-input soft-output (SISO) convolutional decoder, Information is fed back from SISO decoders to SICLCs for interference cancellation. Each SICLC then provides soft information about coded symbols in the form of log-likelihood ratio (LLR), which is used by the corresponding SISO decoder as a priori information about coded symbols. We evaluate the performance of the proposed... 

In this manuscript, we consider using soft-input decoder for internally bandwidth efficient convolutionally coded fiber-optic Code Division Multiple Access (CDMA) communication systems using Optical Orthogonal Codes (OOC). The convolution codes that are used for demonstrating the capabilities of the soft-input decoder are Super Orthogonal Codes (SOC). These codes are near optimal and have a relatively low complexity. We evaluate the upper bounds on the bit error probability of the internally coded fiber-optic CDMA system using the soft-input and hard-input decoders assuming both On-Off Keying (OOK) and Binary Pulse Position Modulation (BPPM) schemes. It is shown that the soft-input decoder... 

Electrocardiogram (ECG) recording of electrical heart activities has a vital diagnostic role in heart diseases. We propose to tackle the problem of arrhythmia detection from ECG signals totally by a deep model that does not need any hand-designed feature or heuristic segmentation (e.g., ad-hoc R-peak detection). In this work, we first segment ECG signals by detecting R-peaks automatically via a convolutional network, including dilated convolutions and residual connections. Next, all beats are aligned around their R-peaks as the most informative section of the heartbeat in detecting arrhythmia. After that, a deep learning model, including both dilated convolution layers and a Long-Short Term... 

Accurate gross tumor volume (GTV) segmentation in esophagus CT images is a critical task in computer aided diagnosis (CAD) systems. However, because of the difficulties raised by the contrast similarity between esophageal GTV and its neighboring tissues in CT scans, this problem has been addressed weakly. In this paper, we present a 3D end-to-end method based on a convolutional neural network (CNN) for this purpose. We leverage design elements from DenseNet in a typical U-shape. The proposed architecture consists of a contractile path and an extending path that includes dense blocks for extracting contextual features and retrieves the lost resolution respectively. Using dense blocks leads to... 

Graph embedding is an important approach for graph analysis tasks such as node classification and link prediction. The goal of graph embedding is to find a low dimensional representation of graph nodes that preserves the graph information. Recent methods like Graph Convolutional Network (GCN) try to consider node attributes (if available) besides node relations and learn node embeddings for unsupervised and semi-supervised tasks on graphs. On the other hand, multi-layer graph analysis has been received attention recently. However, the existing methods for multi-layer graph embedding cannot incorporate all available information (like node attributes). Moreover, most of them consider either... 

Estimation of 3D human pose from a single image is a challenging task because of ambiguities in projection from 3D space to the 2D image plane. A new two-stage deep convolutional neural network-based method is proposed for regressing the distance and angular difference matrices among body joints. Using the angular difference between body joints in addition to the distance between them in articulated objects such as human body can better model the structure of the shapes and increases the modelling capability of the learning method. Experimental results on HumanEva I and Human3.6M datasets show that the proposed method has substantial improvement in the mean per joint position error measure... 

With the ever-growing portion of internet traffic associated with multimedia data and the existence of multiple copies of the same content in various forms, it has become vital to measure the quality of the image and video files. In most cases, without access to the original file, the quality shall be assessed solely based on the available file. Specifically, the challenge of no-reference image quality assessment (NR-IQA) is to predict a quality measure for given images in a consistent manner with human perception of quality. Conventional NR-IQA methods try to fit certain distortion models to a given image and quantify the quality. In practice, however, an image is affected by a combination... 

Segmentation of the Left ventricle (LV) is a crucial step for quantitative measurements such as area, volume, and ejection fraction. However, the automatic LV segmentation in 2D echocardiographic images is a challenging task due to ill-defined borders, and operator dependence issues (insufficient reproducibility). U-net, which is a well-known architecture in medical image segmentation, addressed this problem through an encoder-decoder path. Despite outstanding overall performance, U-net ignores the contribution of all semantic strengths in the segmentation procedure. In the present study, we have proposed a novel architecture to tackle this drawback. Feature maps in all levels of the decoder... 

It has been shown that more accurate signal recovery can be achieved with low-coherence dictionaries in sparse signal processing. In this Letter, the authors extend the low-coherence attribute to receptive fields in convolutional neural networks. A new constrained formulation to train low-coherence convolutional neural network is presented and an efficient algorithm is proposed to train the network. The resulting formulation produces a direct link between the receptive fields of a layer through training procedure that can be used to extract more informative representations from the subsequent layers. Simulation results over three benchmark datasets confirm superiority of the proposed... 

With the recent advancements in deep learning, high performance neural networks have been introduced. These neural networks also can be used to solve similar problems in a transfer learning approach. Recently, several state-of-The-Art Convolutional Neural Networks (CNNs) are proposed for computer vision tasks. On the other hand, in-The-wild No-Reference (Blind) Image Quality Assessment (NR-IQA) problem is known as a challenging human perceptual problem. In this paper, a transfer learning approach is used to solve the problem of in-The-wild NR-IQA. With a few training times, the proposed neural network exceeds all the previous methods which are not using deep neural networks. Further, the... 

We prove an inequality for the pth power of the norm of a stochastic convolution integral in a Hilbert space. The inequality is stronger than analogous inequalities in the literature in the sense that it is pathwise and not in expectation  

Semilinear stochastic evolution equations with multiplicative Lévy noise are considered. The drift term is assumed to be monotone nonlinear and with linear growth. Unlike other similar works, we do not impose coercivity conditions on coefficients. We establish the continuous dependence of the mild solution with respect to initial conditions and also on coefficients. As corollaries of the continuity result, we derive sufficient conditions for asymptotic stability of the solutions, we show that Yosida approximations converge to the solution and we prove that solutions have Markov property. Examples on stochastic partial differential equations and stochastic delay differential equations are... 

Semilinear stochastic evolution equations with multiplicative Poisson noise and monotone nonlinear drift in Hilbert spaces are considered. The coefficients are assumed to have linear growth. We do not impose coercivity conditions on coefficients. A novel method of proof for establishing existence and uniqueness of the mild solution is proposed. Examples on stochastic partial differential equations and stochastic delay differential equations are provided to demonstrate the theory developed. © 2017 Iranian Mathematical Society