فصلنامه پردازش علائم و داده ها
سال هفدهم شماره 2 (پیاپی 44، تابستان 1399)

Supervisory control and data acquisition (SCADA) system monitors and controls industrial processes in critical infrastructures (CIs) and plays the vital role in maintaining the reliability of CIs such as power, oil, and gas system. In fact, SCADA system refers to the set of control process, which measures and monitors sensors in remote substations from a control center. These sensors usually have a type of automated response capability when a certain criteria is met. When an abnormal system status occurs, an alarm signal is raised in control center and as a result the operator will be notified. In this way, all normal and abnormal system statuses are monitored in control center. In CI’s application, since several substation resources and their related sensors are too high (because the CI’s grid is often large, complex and wide), the number of alarms is very high. It gets worse when the operator mistakes and as a result, cascading alarms are flooded. In this condition, the rate of raising alarms may be more than clearing them. In SCADA system, alarm clearing is one of the main duties of operators. When an alarm is raised in control center, the operator should clear it as soon as possible. However, the recent reports confirm the poor alarm clearing causes accidents in the SCADA system. As any operator mistake can increase the number of alarms and jeopardize the system reliability, alarms processing and decision-making for clearing them are a stressful and time-consuming for the SCADA operators. In a large and complex CI such as power system, when operators are overwhelmed by the system alarms, they may take wrong decisions and even ignore alarms. Alarm flooding, lots of operator’s workload and his/her fatigue as a result, are the main causes of operator’s mistake. If generating of an alarm in a remote substation is denoted as an operational cycle in an SCADA system until clearing it by the operator in control center, the aim of this paper is modeling the operational cycle by using colored petri nets. The proposed model is based on a general approach which alarm messages are integrated with the operator’s commands. Of course, the model focuses on generating of alarms by substation resources. To verify the proposed model, a real data set of power system of Iran is used and to demonstrate the potential of the proposed model some scenarios about operator’ workload and alarm flooding are simulated.

Recently, orthogonal frequency division multiplexing (OFDM) has been extensively used in communications systems to resist channel impairments in frequency selective channels. OFDM is a multicarrier transmission technology in wireless environment that use a large number of orthogonal subcarriers to transmit information. OFDM is one of the most important blocks in digital video broadcast-terrestrial (DVB-T) system. The goal of this paper is comparing the methods of interpolation in OFDM system that not used channel statistics information. Therefore, we used pilots for obtaining the information of channel, and by the method of estimation without use of channel statistics information, the channel primary frequency response estimated in pilot’s frequencies. Pilots, channel estimation and interpolation methods are key roles in the OFDM block. The number of pilots are different in the OFDM symbol for different pilot patterns.  In this article, we proposed three pilot patterns to improve DVB-T system performance. Our criteria for this purpose are error probability, calculation time, and the number of pilots. We have tested the performance improvement by using two-dimensional (2D) interpolation methods. Obviously, we do not obtain all of our requests and requirements via one pilot pattern. For example, the error may be decreases, but the number of pilots is increased. Therefore, we must select the pilot pattern that achieve the most important goal for us. We have applied six interpolation methods, for 2D interpolation, such as linear, nearest-neighbor, spline, cubic Hermite, cosine and low pass interpolations. We have compared three proposed pilot patterns with the conventional DVB-T pilot pattern in four different channels. In each channel, we have tested 30 interpolation methods. The applied channels are OFDM system with AWGN noise, OFDM system with AWGN noise and Rayleigh fading, DVB-T system with AWGN noise and DVB-T system with AWGN noise and Rayleigh fading. We observed that the best performance happens when we use linear interpolation in the first dimension and cosine interpolation in the second dimension of 2D interpolation. In addition, the worst performance will be happened when Nearest-neighbor interpolation is used in the second dimension of 2D interpolation. In the last step, we compared the proposed pilot patterns with the conventional DVB-T pilot pattern in 2D interpolation method that it leads to better performance in DVB-T system. We observed that the proposed pilot patterns have better performance than the conventional DVB-T pilot pattern. In the DVB-T, movement and velocity are very important and considered in this research. In the second step using DVB-T pilot pattern, we compared 2D interpolation methods in some different Doppler frequencies. Simulation results show that at 3 Doppler frequencies, i.e. 0, 30, 150Hz, the proposed schemes with a linear interpolation has better performance than the conventional method in the DVB-T systems.

Designing a secure software is one of the major phases in developing a robust software. The McGraw life cycle, as one of the well-known software security development approaches, implements different touch points as a collection of software security practices. Each touch point includes explicit instructions for applying security in terms of design, coding, measurement, and maintenance of software. Developers are able to provide secure and robust software by applying such touch points. In this paper, we introduce a secure and robust approach to map McGraw cycle to RUP methodology, named RUPST. The traditional form of RUP methodology is revised based on the proposed activities for software security. RUPST adds activities like security requirements analysis, abuse case diagrams, risk-based security testes, code review, penetration testing, and security operations to the RUP disciplines. In this regard, based on RUP disciplines, new touch points of software security are presented as a table. Also, RUPST adds new roles such as security architect and requirement analyzer, security requirement designer, code reviewer and penetration tester which are presented in the form of a table along with responsibilities of each role. This approach introduces new RUP artifacts for disciplines and defines new roles in the process of secure software design. The offered artifacts by RUPST include security requirement management plan, security risk analysis model, secure software architecture document, UMLSec model, secure software deployment model, code review report, security test plan, security testes procedures, security test model, security test data, penetration report, security risks management document, secure installation and configuration document and security audit report. We evaluate the performance of the RUPST in real software design process in comparison to other secure software development approaches for different security aspects. The results demonstrate the efficiency of   the proposed methodology in developing of a secure and robust software.

Image representation is a crucial problem in image processing where there exist many low-level representations of image, i.e., SIFT, HOG and so on. But there is a missing link across low-level and high-level semantic representations. In fact, traditional machine learning approaches, e.g., non-negative matrix factorization, sparse representation and principle component analysis are employed to describe the hidden semantic information in images, where they assume that the training and test sets are from same distribution. However, due to the considerable difference across the source and target domains result in environmental or device parameters, the traditional machine learning algorithms may fail. Transfer learning is a promising solution to deal with above problem, where the source and target data obey from different distributions. For enhancing the performance of model, transfer learning sends the knowledge from the source to target domain. Transfer learning benefits from sample reweighting of source data or feature projection of domains to reduce the divergence across domains. Sparse coding joint with transfer learning has received more attention in many research fields, such as signal processing and machine learning where it makes the representation more concise and easier to manipulate. Moreover, sparse coding facilitates an efficient content-based image indexing and retrieval. In this paper, we propose image classification via Sparse Representation and Subspace Alignment (SRSA) to deal with distribution mismatch across domains in low-level image representation. Our approach is a novel image optimization algorithm based on the combination of instance-based and feature-based techniques. Under this framework, we reweight the source samples that are relevant to target samples using sparse representation. Then, we map the source and target data into their respective and independent subspaces. Moreover, we align the mapped subspaces to reduce the distribution mismatch across domains. The proposed approach is evaluated on various visual benchmark datasets with 14 experiments. Comprehensive experiments demonstrate that SRSA outperforms other latest machine learning and domain adaptation methods with significant difference.

Cystic fibrosis (CF) is the most common autosomal recessive disorder in white skinned individuals. Chronic lung infection is the main cause of mortality in this disease. Approximately 60–75 % of adult CF patients frequently suffer from Pseudomonas aeruginosa (PA) infection that is strongly associated with inflammation, lung destruction, and increased mortality. Therefore, CF patients should be followed up by physicians to diagnose infection in the primary stage, start treatment, and reduce the risk of chronic infection. Although sputum culture is the gold standard for diagnosis of PA infections, a rapid and accurate diagnostic method can facilitate early initiation of appropriate therapy and easy monitoring of the condition. The aim of this study was to diagnose CF patients with infection using their lung sound. In this study, the symmetry of frequency information in right and left lung was investigated in CF patients with positive sputum culture results, negative sputum culture results‎, and patients who underwent treatment with antibiotics. Respiretorysounds were acquired from 34 CF patients (16 female, 18 male) who were being ‎followed-up at the Pediatric Respiratory and Sleep Medicine Research Center of Childrenchr('39')s ‎Medical Center. The patient selection was based on their sputum microbiology culture. The selection ‎category was as follows: 12 patients with normal flora culture results and 11 patients with PA ‎infection. Also, respiratory sounds of 11 patients were recorded one month after antibiotic treatment and they used to investigate the effectiveness of the proposed method. In the preprocessing step, cardiac sound was removed, respiratory sound cycles were separated and the signals were divided into 64 milisecond frame and 15 features were extracted from each frame. Differences between these features were computed between right and left lungs for early, middle and late section of the respiratory cycle using the new proposed feature. Then, the best group of features was selected by applying Genetic Algorithm. The selected group of features was fed into Support Vector Machine, K Nearest Neighbor and Naïve Bayesian classifier. Also, an Ensemble classifier was examined. The best result was obtained by Ensemble classifier that diagnosed infection by the accuracy of 91.3% and differentiates a group of CF patients with infection from CF patients who underwent treatment with an accuracy of 90.9%. This study describes a novel method of infection detection in CF patients based only on respiratory sound analysis. The proposed method is a simple and available way for early diagnosis of infection and initiating therapeutic strategies.

Due to some physiological and physical limitations in the brain and the eye, the human visual system (HVS) is unable to perceive some changes in the visual signal whose range is lower than a certain threshold so-called just-noticeable distortion (JND) threshold. Visual attention (VA) provides a mechanism for selection of particular aspects of a visual scene so as to reduce the computational load on the brain. According to the current knowledge, it is believed that VA is driven by “visual saliency”. In a visual scene, a region is said to be visually salient if it possess certain characteristics, which make it stand out from its surrounding regions and draw our attention to it. In most existing researches for estimating the JND threshold, the sensitivity of the HVS has been consider the same throughout the scene and the effects of visual attention (caused by visual saliency) which have been ignored. Several studies have shown that in salient areas that attract more visual attention, visual sensitivity is higher, and therefore the JND thresholds are lower in those points and vice versa. In other words, visual saliency modulates JND thresholds. Therefore, considering the effects of visual saliency on the JND threshold seems not only logical but also necessary. In this paper, we present an improved non-uniform model for estimating the JND threshold of images by considering the mechanism of visual attention and taking advantage of visual saliency that leads to non-uniformity of importance of different parts of an image. The proposed model, which has the ability to use any existing uniform JND model, improves the JND threshold of different pixels in an image according to the visual saliency and by using a non-linear modulation function. Obtaining the parameters of the nonlinear function through an optimization procedure leads to an improved JND model. What make the proposed model efficient, both in terms of computational simplicity, accuracy, and applicability, are: choosing nonlinear modulation function with minimum computational complexity, choosing appropriate JND base model based on simplicity and accuracy and also Computational model for estimating visual saliency  that accurately determines salient areas, Finally, determine the Efficient cost function and solve it by determining the appropriate  objective Image Quality Assessment. To evaluate the proposed model, a set of objective and subjective experiments were performed on 10 selected images from the MIT database. For subjective experiment, A Two Alternative Forced Choice (2AFC) method was used to compare subjective image quality and for objective experiment SSIM and IWSSIM was used. The obtained experimental results demonstrated that in subjective experiment the proposed model achieves significant superiority than other existing models and in objective experiment, on average, outperforms the compared models. The computational complexity of proposed model is also analyzed and shows that it has faster speed than compared models.

Clustering algorithms are highly dependent on different factors such as the number of clusters, the specific clustering algorithm, and the used distance measure. Inspired from ensemble classification, one approach to reduce the effect of these factors on the final clustering is ensemble clustering. Since weighting the base classifiers has been a successful idea in ensemble classification, in this paper we propose a method to use weighting in the ensemble clustering problem. The accuracies of base clusterings are estimated using an algorithm from crowdsourcing literature called agreement/disagreement method (AD). This method exploits the agreements or disagreements between different labelers for estimating their accuracies. It assumes different labelers have labeled a set of samples, so each two persons have an agreement ratio in their labeled samples. Under some independence assumptions, there is a closed-form formula for the agreement ratio between two labelers based on their accuracies. The AD method estimates the labelers’ accuracies by minimizing the difference between the parametric agreement ratio from the closed-form formula and the agreement ratio from the labels provided by labelers. To adapt the AD method to the clustering problem, an agreement between two clusterings are defined as having the same opinion about a pair of samples. This agreement can be as either being in the same cluster or being in different clusters. In other words, if two clusterings agree that two samples should be in the same or different clusters, this is considered as an agreement. Then, an optimization problem is solved to obtain the base clusterings’ accuracies such that the difference between their available agreement ratios and the expected agreements based on their accuracies is minimized. To generate the base clusterings, we use four different settings including different clustering algorithms, different distance measures, distributed features, and different number of clusters. The used clustering algorithms are mean shift, k-means, mini-batch k-means, affinity propagation, DBSCAN, spectral, BIRCH, and agglomerative clustering with average and ward metrics. For distance measures, we use correlation, city block, cosine, and Euclidean measures. In distributed features setting, the k-means algorithm is performed for 40%, 50%,…, and 100% of randomly selected features. Finally, for different number of clusters, we run the k-means algorithm by k equals to 2 and also 50%, 75%, 100%, 150%, and 200% of true number of clusters. We add the estimated weights by the AD algorithm to two famous ensemble clustering methods, i.e., Cluster-based Similarity Partitioning Algorithm (CSPA) and Hyper Graph Partitioning Algorithm (HGPA). In CSPA, the similarity matrix is computed by taking a weighted average of the opinions of different clusterings. In HGPA, we propose to weight the hyperedges by different values such as the estimated clustering accuracies, size of clusters, and the silhouette of clusterings. The experiments are performed on 13 real and artificial datasets. The reported evaluation measures include adjusted rand index, Fowlkes-Mallows, mutual index, adjusted mutual index, normalized mutual index, homogeneity, completeness, v-measure, and purity. The results show that in the majority of cases, the proposed weighted-based method outperforms the unweighted ensemble clustering. In addition, the weighting is more effective in improving the HGPA algorithm than CSPA. For different weighting methods proposed for HGPA algorithm, the best average results are obtained when we use the accuracies estimated by the AD method to weight the hyperedges, and the worst results are obtained when using the normalized silhouette measure for weighting. Finally, among different methods for generating base clusterings, the best results in weighted HGPA are obtained when we use different clustering algorithms to come up with different base clusterings.

Deep understanding of molecular biology has allowed emergence of new technologies like DNA decryption.  On the other hand, advancements of molecular biology have made manipulation of genetic systems simpler than ever; this promises extraordinary progress in biological, medical and biotechnological applications.  This is not an unrealistic goal since genes which are regulated by gene regulatory networks (GRNs) are the core governors of life processes at the molecular level. In fact, manipulation of GRNs would be the ultimate strategy for optimal purposeful control of cell’s life.  GRNs are in charge of regulating the amounts of all the inter-cellular as well as intra-cellular molecular species produced all the time in all living organisms.  Manipulation of a GRN requires comprehensive knowledge about nodes and interconnections.  This paper deals with both aspects in networks having more than fifty nodes.  In the first part of the paper, restrictions of probabilistic models in modeling node behavior are discussed, i.e.: 1) unfeasibility of reliably predicting the next state of GRN based on its current state, 2) impossibility of modelling logical relations among genes, and 3) scarcity of biological data needed for model identification.  These findings which are supported by arguments from probability theory suggest that probabilistic models should not be used for analysis and prediction of node behavior in GRNs.  Next part of the paper focuses on models of GRN structure.  It is shown that the use of multi-tree models for structure for GRN poses severe limitations on network behavior, i.e. 1) increase in signal entropy while passing through the network, 2) decrease in signal bandwidth while passing through the network, and 3) lack of feedback as a key element for oscillatory and/or autonomous behavior (a requirement for any biological network).  To demonstrate that, these restrictions are consequences of model selection, we use information theoretic arguments.  At the last and the most important part of the paper we look into the gene perturbation experiments from a network-theoretic perspective to show that multi-perturbation experiments are not as informative as assumed so far.  A generally accepted belief among researches states that multi-perturbation experiments are more informative than single-perturbation ones, i.e., multiple simultaneously applied perturbations provide more information than a single perturbation.  It is shown that single-perturbation experiments are optimal for identification of network structure, provided the ultimate goal is to discover correct subnet structures.

In this study, a Brain-Computer Interface (BCI) in Silent-Talk application was implemented. The goal was an electroencephalograph (EEG) classifier for three different classes including two imagined words (Man and Red) and the silence. During the experiment, subjects were requested to silently repeat one of the two words or do nothing in a pre-selected random order. EEG signals were recorded by a 14 channel EMOTIV wireless headset. Two combinations of features and classifiers were used: Discrete Wavelet Transform (DWT) features with Support Vector Machine (SVM) classifier and Principle Component Analysis (PCA) features with a Minimum-Distance classifier. Both combinations were capable of discriminating between the three classes much better than the chance level (33.3%), none of them was reliable and accurate enough for a real application though. The first method (DWT+SVM) showed better results. In this case, feature set was D2, D3, D4 and A4 coefficients of 4-level DWT decomposition of the EEG signals, roughly corresponding to major frequency bands (Delta, Theta, Alpha and Beta) of these signals. Three binary SVM machines were used. Each machine was trained to classify between two of the three classes, namely Man/Red, Man/Silence or Red/Silence. Majority Selection Rule was used to determine final class. Once two of these classifiers presented the true class, a win (correct classification) was counted, otherwise a loss (false classification) was considered. Finally, Monte-Carlo Cross Validation showed an overall performance of about 56.8% correct classification which is comparable with the results reported for similar experiments.

Coreference resolution is an advanced issue in natural language processing. Nowadays, due to the extension of social networks, TV channels, news agencies, the Internet, etc. in human life, reading all the contents, analyzing them, and finding a relation between them require time and cost. In the present era, text analysis is performed using various natural language processing techniques, one of the challenges in this field is the low accuracy in detecting name entitieschr('39') reference, which detection process has been named as coreference resolution. Coreference resolution is finding all expressions that refer to a name entity, and two expressions are coreference together when these expressions located in the same coreference cluster.      Coreference resolution could be used in many natural language processing tasks such as question answering, text summarization, machine translation, information extraction, etc. Coreference resolution methods are into two main categories; machine learning and rule-based approaches. In the rule-based approaches for detecting coreferences, a set of rich rule ordinary which written by a specialist is execued. These methods are quick, but these are language-dependent and necessary written to each language firstly again by a specialist. The machine learning method divides into supervised and unsupervised methods, in a supervised approach, it is require to have data labeled by a specialist. Coreference resolution included three main phases: named entities recognition, features extraction of name entities, and analyzes the coreferences, in which the primary phase is feature extraction. After corpus creation, name entities should be recognized in the corpus. This step depends on a corpus, in some corpora entities named as golden data, in this paper, we used RCDAT corpus, which determined name entities itself. After the name entities recognition phase, the mention pairs are determined, and the features are extracted. The proposed method uses two categories of the features: the first is word embedding vector, the second is handcrafted features, which are the distance between the mentions, head matching, gender matching, etc. This paper used a deep neural network to train the features extracted, in the analyze coreferences phase a Feed Forward Neural Network (FFNN) is trained by the candidate mention pairs (extracted features from them) and their labels (coreference / non-coreference or 1/0) so that the trained FFNN assigns a probability (between 0 and 1) to any given mention pair. Then used the graph technique with a threshold level to determine different or compatible name entities in the coreference resolution cluster.  This step creates the graph by using the extracted mention pairs from the previous step. In this graph, nodes are the mention pairs that are clustered by using the agglomerative hierarchical clustering algorithm inorder to locate similar mention pairs in a group. The resulting clusters are considered as coreference resolution chains. In this paper, RCDAT Persian language corpus is used for training the proposed coreference resolution approach and for testing the Uppsala Persian language dataset which is used and in the calculation of the accurate of system, different tools have been taken for features extraction which each of them effects on the accuracy of the whole system. The corpora, tools, and methods used in the system are standard. They are quite comparable to the ACE and Ontonotes corpora and tools used at the same time in the coreference resolution algorithm.  The results of the improvements proposed method (F1 = 62.09) is expressed in the text of the paper.