Vice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Exploiting Sparse Representation for Sleep Stage Classification Using Electroencephalogram SignalExploiting Sparse Representation for Sleep Stage Classification Using Electroencephalogram Signal111917610.22034/jasp.2019.9176FAB.AzadianDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranT.Yousefi RezaiiDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranS.MeshginiDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranJournal Article20190406In this paper, sparse representation of EEG signal is used to automatically classify sleep stages. In this regard, two general sparse representation trends are proposed to classify 4-class sleep stages. The first proposed method is based on sparse principal component analysis (SPCA) which uses different features including time, frequency, and time-frequency features applied to support vector machine (SVM) classifier. The second proposed method is based on sparse representation-based classifier (SRC) which uses orthogonal matching pursuit (OMP) algorithm to obtain sparse coding of the EEG signal. In order to evaluate the effectiveness of the proposed algorithms, their performance is compared with the conventional SVM classification based on PCA method using time, frequency, and time-frequency features. The study is carried out on EEG signal from Physionet international database. Simulation results show on the average 8.36% and 8.26% improvement of the first proposed method in terms of classification accuracy compared to the PCA and deep learning methods, respectively, while the second proposed method has achieved the running time of 118% and 72% faster than the existing PCA and deep learning methods, respectively.In this paper, sparse representation of EEG signal is used to automatically classify sleep stages. In this regard, two general sparse representation trends are proposed to classify 4-class sleep stages. The first proposed method is based on sparse principal component analysis (SPCA) which uses different features including time, frequency, and time-frequency features applied to support vector machine (SVM) classifier. The second proposed method is based on sparse representation-based classifier (SRC) which uses orthogonal matching pursuit (OMP) algorithm to obtain sparse coding of the EEG signal. In order to evaluate the effectiveness of the proposed algorithms, their performance is compared with the conventional SVM classification based on PCA method using time, frequency, and time-frequency features. The study is carried out on EEG signal from Physionet international database. Simulation results show on the average 8.36% and 8.26% improvement of the first proposed method in terms of classification accuracy compared to the PCA and deep learning methods, respectively, while the second proposed method has achieved the running time of 118% and 72% faster than the existing PCA and deep learning methods, respectively.https://jasp.tabrizu.ac.ir/article_9176_b1d676ee6de407a1ab5c16401562bfbb.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Vessel Extraction of Retinal Images of Diabetic Retinopathy Using A Morphology-based AlgorithmVessel Extraction of Retinal Images of Diabetic Retinopathy Using A Morphology-based Algorithm1323917710.22034/jasp.2019.9177FAZ.Asgharzadeh BonabDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranS.MeshginiDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranJournal Article20190314Diabetes is a common disease in the world. The first member that is usually damaged is the eye. Diabetic retinopathy is a diabetic disorder and occurs due to changes in the blood vessels of the retina. Extracting blood vessels is initial step for diagnosis of retina problems. Imaging of Retina needs some special cameras called fundus. It is a digital camera that captures retina images ad is capable to save them. The purpose of this paper is to provide a method for the diagnosis of blood vessels based on morphology on retina images. After converting a color image to a gray scale one and improving the quality, morphological operators are used to remove the optical disk from the image. Then, blood vessels are extracted from the retina image by two different methods. Combining these two methods gives more detailed results. Possible noise is then removed using median filters. Finally, the results are combined and the blood vessels are extracted. The proposed algorithm has been evaluated over the images from the Drive database. The experimental results shows the effectiveness of our proposed method. The average result of specificity, sensitivity and accuracy are 0.98, 0.751 and 0.960, respectively.Diabetes is a common disease in the world. The first member that is usually damaged is the eye. Diabetic retinopathy is a diabetic disorder and occurs due to changes in the blood vessels of the retina. Extracting blood vessels is initial step for diagnosis of retina problems. Imaging of Retina needs some special cameras called fundus. It is a digital camera that captures retina images ad is capable to save them. The purpose of this paper is to provide a method for the diagnosis of blood vessels based on morphology on retina images. After converting a color image to a gray scale one and improving the quality, morphological operators are used to remove the optical disk from the image. Then, blood vessels are extracted from the retina image by two different methods. Combining these two methods gives more detailed results. Possible noise is then removed using median filters. Finally, the results are combined and the blood vessels are extracted. The proposed algorithm has been evaluated over the images from the Drive database. The experimental results shows the effectiveness of our proposed method. The average result of specificity, sensitivity and accuracy are 0.98, 0.751 and 0.960, respectively.https://jasp.tabrizu.ac.ir/article_9177_b41ff156db1669a5b8b76b7973af1d8e.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Epileptic Seizure Prediction Using Heart Rate Variability Signal AnalysisEpileptic Seizure Prediction Using Heart Rate Variability Signal Analysis2533917810.22034/jasp.2019.9178FAM.PoudinehDepartment of Biomedical Engineering, K. N. Toosi University of Technology, Tehran, IranM.MohebbiDepartment of Biomedical Engineering, K. N. Toosi University of Technology, Tehran, IranK.GharagozliNeuroscience Group, Shahid Beheshti University of Medical Sciences, Tehran, IranJournal Article20170927Epilepsy is a neural disorder with unknown nature and epileptic patients suffer from the consequences of unexpected seizures. In this paper, we proposed a new method to predict epileptic seizures using heart rate variability (HRV) signal analysis. During preictal period of epilepsy, increasing in nervous activities of neurons affects the autonomic nervous system that disturbs heart rates. Therefore, epileptic seizures can be predicted through HRV monitoring. In our method, we extracted 12 features of HRV signal from different domains: time, frequency, time-frequency and non-linear domain. We used Multivariate Statistical Process Control (MSPC) algorithm for anomaly detection which is able to detect anomalies that cannot be detected by monitoring each variable independently. This algorithm has been applied to the clinical data collected from 17 patients. The obtained results demonstrated that the proposed method can predict seizure onset with an accuracy of 88.2%. The proposed HRV-based seizure prediction algorithm is more promising than the conventional EEG-based methods from the viewpoint of practical use.Epilepsy is a neural disorder with unknown nature and epileptic patients suffer from the consequences of unexpected seizures. In this paper, we proposed a new method to predict epileptic seizures using heart rate variability (HRV) signal analysis. During preictal period of epilepsy, increasing in nervous activities of neurons affects the autonomic nervous system that disturbs heart rates. Therefore, epileptic seizures can be predicted through HRV monitoring. In our method, we extracted 12 features of HRV signal from different domains: time, frequency, time-frequency and non-linear domain. We used Multivariate Statistical Process Control (MSPC) algorithm for anomaly detection which is able to detect anomalies that cannot be detected by monitoring each variable independently. This algorithm has been applied to the clinical data collected from 17 patients. The obtained results demonstrated that the proposed method can predict seizure onset with an accuracy of 88.2%. The proposed HRV-based seizure prediction algorithm is more promising than the conventional EEG-based methods from the viewpoint of practical use.https://jasp.tabrizu.ac.ir/article_9178_1a0ddea7c248c6ab7ddcd230dcbc9823.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Detection of Epilepsy in Electroencephalographic (EEG) Signals Based on Global Wavelet Spectrum (GWS) Using Support Vector Machine (SVM)Detection of Epilepsy in Electroencephalographic (EEG) Signals Based on Global Wavelet Spectrum (GWS) Using Support Vector Machine (SVM)3543917910.22034/jasp.2019.9179FAF.HasanzadehDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranS.MeshginiDepartment of Biomedical Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, IranJournal Article20190212Approximately one percent of the world's population suffers from epilepsy. The first stage of epilepsy treatment is timely and correct diagnosis. One of the ways to diagnose epilepsy is to accurately analyze EEG signals. There are various features to diagnose the disease from a signal such as the signal amplitude. In this paper, a new method for the diagnosis of epilepsy is presented by examining the time-frequency information of the EEG signal in people with seizure-free seizure syndrome and healthy people. Initially, the Global Wavelet Spectrum (GWS) feature of the EEG signal was extracted. To interpret this Spectrum in frequency bands, EEG signals decompose to five levels by continuous wavelet transform. Then, by applying this feature, a Support vector machine-based classifier was used to diagnose epilepsy. The results of the analysis provided a significant difference in the separation of the individual based on the brain signal. The proposed method compared to the previous methods, can classify epilepsy and intact signals with 100% accuracy. It was also observed that the dominant (GWS) values for the signals selected from patients with epilepsy in the delta and theta frequency band are discussed.Approximately one percent of the world's population suffers from epilepsy. The first stage of epilepsy treatment is timely and correct diagnosis. One of the ways to diagnose epilepsy is to accurately analyze EEG signals. There are various features to diagnose the disease from a signal such as the signal amplitude. In this paper, a new method for the diagnosis of epilepsy is presented by examining the time-frequency information of the EEG signal in people with seizure-free seizure syndrome and healthy people. Initially, the Global Wavelet Spectrum (GWS) feature of the EEG signal was extracted. To interpret this Spectrum in frequency bands, EEG signals decompose to five levels by continuous wavelet transform. Then, by applying this feature, a Support vector machine-based classifier was used to diagnose epilepsy. The results of the analysis provided a significant difference in the separation of the individual based on the brain signal. The proposed method compared to the previous methods, can classify epilepsy and intact signals with 100% accuracy. It was also observed that the dominant (GWS) values for the signals selected from patients with epilepsy in the delta and theta frequency band are discussed.https://jasp.tabrizu.ac.ir/article_9179_5ae4ea9bdfff2e700e65679d6e589c8c.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Simultaneous Energy Harvesting and Information Processing in Wireless Communications Using Multiple Relays with Multiple Antennas Considering Various Locations of RelaysSimultaneous Energy Harvesting and Information Processing in Wireless Communications Using Multiple Relays with Multiple Antennas Considering Various Locations of Relays4555918010.22034/jasp.2019.9180FAA.ZahediDepartment of Electrical Engineering, Kermanshah University of Technology, Kermanshah, Iran0000-0003-4809-984XJournal Article20170719Wireless networks suffer from battery discharging specially in cooperative communications, where multiple energy constrained relays are used. To overcome this problem, energy harvesting from RF signals is used to charge the node battery. These intermediate nodes have the ability to harvest energy from the source signal and use the harvested energy to transmit information to the destination. In fact, the node tries to harvest energy and then transmit the data to destination. Division of energy harvesting and data transmission periods can be done in two different protocols, namely time switching based relaying protocol (TSR) and power splitting based relaying protocol (PSR). These two protocols can also be applied in delay-limited and delay-tolerant transmission systems. The previous works have assumed a single relay for energy harvesting; However, in this paper, the proposed method is concentrated on improving the outage probability and throughput using multiple antennas in each relay node instead of using single antenna. Also the optimum location of multiple relays is discussed and its effect on throughput of the system is mainly considered. According to our simulation results, when multi-antenna relays are used, ability of energy harvesting is increased and thus system performance will be improved to a great extent. MRC selection relay scheme is used when the destination chooses a group of relays and antennas satisfying the required SNR.Wireless networks suffer from battery discharging specially in cooperative communications, where multiple energy constrained relays are used. To overcome this problem, energy harvesting from RF signals is used to charge the node battery. These intermediate nodes have the ability to harvest energy from the source signal and use the harvested energy to transmit information to the destination. In fact, the node tries to harvest energy and then transmit the data to destination. Division of energy harvesting and data transmission periods can be done in two different protocols, namely time switching based relaying protocol (TSR) and power splitting based relaying protocol (PSR). These two protocols can also be applied in delay-limited and delay-tolerant transmission systems. The previous works have assumed a single relay for energy harvesting; However, in this paper, the proposed method is concentrated on improving the outage probability and throughput using multiple antennas in each relay node instead of using single antenna. Also the optimum location of multiple relays is discussed and its effect on throughput of the system is mainly considered. According to our simulation results, when multi-antenna relays are used, ability of energy harvesting is increased and thus system performance will be improved to a great extent. MRC selection relay scheme is used when the destination chooses a group of relays and antennas satisfying the required SNR.https://jasp.tabrizu.ac.ir/article_9180_4e71edc2243e3c5f970488fd6e832c83.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Blind Attack on Watermarked Images Based on Segmentation and Non-Local SimilarityBlind Attack on Watermarked Images Based on Segmentation and Non-Local Similarity5765918110.22034/jasp.2019.9181FAS. H.SoleymaniDepartment of Computer Engineering, Ferdowsi University of Mashhad, Mashhad, IranA. H.TaheriniaDepartment of Computer Engineering, Ferdowsi University of Mashhad, Mashhad, IranJournal Article20170927In this paper, a new attack is proposed on blind image watermarking, which is able to destroy the watermark embedded in an image using quantization and dither modulation embedding methods. This attack is consist of three main steps: The first two steps are segmentation and finding more similar regions of image and the third step is swapping similar regions with each other. One of the main idea in the proposed method is the algorithm of finding similar segments, which are far from each other and also are in indistinctive locations. This attack does not need to know the embedding and extraction algorithms and their parameters which are used. Therefore, this method is a blind attack. The results of this algorithm and its comparison with other intentional and unintentional attacks show that it can destroy embedded watermark properly and can preserve the quality of watermarked image. The value of NC metric is less than 0.4 and the value of PSNR metric is about 39 dB between the watermarked image and the attacked image.In this paper, a new attack is proposed on blind image watermarking, which is able to destroy the watermark embedded in an image using quantization and dither modulation embedding methods. This attack is consist of three main steps: The first two steps are segmentation and finding more similar regions of image and the third step is swapping similar regions with each other. One of the main idea in the proposed method is the algorithm of finding similar segments, which are far from each other and also are in indistinctive locations. This attack does not need to know the embedding and extraction algorithms and their parameters which are used. Therefore, this method is a blind attack. The results of this algorithm and its comparison with other intentional and unintentional attacks show that it can destroy embedded watermark properly and can preserve the quality of watermarked image. The value of NC metric is less than 0.4 and the value of PSNR metric is about 39 dB between the watermarked image and the attacked image.https://jasp.tabrizu.ac.ir/article_9181_2aa6e24b76419c389ee7f5bb2ab16308.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Evaluation of Segmentation and Bias Field Correction in MR Brain Images Using Level Set and Multiplicative Intrinsic Component Optimization MethodsEvaluation of Segmentation and Bias Field Correction in MR Brain Images Using Level Set and Multiplicative Intrinsic Component Optimization Methods6775918210.22034/jasp.2019.9182FAA.Alipour SifarFaculty of Electrical Engineering, Islamic Azad University, Science & Research Branch, Tabriz, IranM.ShamsiFaculty of Biomedical Engineering, Sahand University of Technology, Tabriz, IranJournal Article20170510Segmentation of brain MR images is a major issue in medical image processing computations. In these images, segmentation is failed by the existence of internal artifact which is called intensity inhomogeneity due to the existence of overlap effect among brain tissue intensities which often causes false classification of brain tissues. In this paper, two suggested methods for segmentation and bias field correction arises, which these images are implemented through the level set (LSM) and multiplicative intrinsic component (MICO) algorithms. Methods outlined in this article include: bias field correction of the human brain MR images by one of these algorithms and segmentation by other algorithm and vice versa. Quantitative and qualitative analysis on the final results showed, accuracy above 90% for the area containing the CSF using the MICO algorithm as well as the areas WM and GM by LSM algorithm. These results can be used to select efficient algorithm to correct the bias field and segmenting each area, separately.Segmentation of brain MR images is a major issue in medical image processing computations. In these images, segmentation is failed by the existence of internal artifact which is called intensity inhomogeneity due to the existence of overlap effect among brain tissue intensities which often causes false classification of brain tissues. In this paper, two suggested methods for segmentation and bias field correction arises, which these images are implemented through the level set (LSM) and multiplicative intrinsic component (MICO) algorithms. Methods outlined in this article include: bias field correction of the human brain MR images by one of these algorithms and segmentation by other algorithm and vice versa. Quantitative and qualitative analysis on the final results showed, accuracy above 90% for the area containing the CSF using the MICO algorithm as well as the areas WM and GM by LSM algorithm. These results can be used to select efficient algorithm to correct the bias field and segmenting each area, separately.https://jasp.tabrizu.ac.ir/article_9182_15319086efd84948c8f0dc5fb2cae28e.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522A Digital Design and Hardware Architecture for the HPRF Radar Signal ProcessorA Digital Design and Hardware Architecture for the HPRF Radar Signal Processor7781918310.22034/jasp.2019.9183FAM. J.FirouziAeronautical University of Shahid Sattari, Tehran, IranH. R.Dalili OskoueiAeronautical University of Shahid Sattari, Tehran, IranR.Fatemi MofradFaculty of Electrical Engineering, Industrial University of Malek-e-Ashtar, Tehran, IranJournal Article20170905Today, the design and use of high-frequency repeater radars are very common because of the advantages of high power transmission and coping with extended clutters and jamming. However, it is always the design and implementation of digital processors that can handle system parameters such as the wide dynamic range above the input signal of these radar coverage has been challenging. In this paper, a method for digital design and determination of the hardware architecture of a high repetition frequency radar signal processor based on the use of software provided by Xilinx XSG It's easier to design and develop The FPGA-based chip-based is provided in the MATLAB Simulink software is presented, this method is based on the use of Software provided by Xilinx Inc. The results of hardware simulation and comparison of output blocks of processing blocks with the output of the analog blocks of the typical radar and comparison with analog digital combined hardware of general radars represent improvement dynamic range of input at least 70 dB and low weight of this processor for a radar with high pulse repetition frequency.Today, the design and use of high-frequency repeater radars are very common because of the advantages of high power transmission and coping with extended clutters and jamming. However, it is always the design and implementation of digital processors that can handle system parameters such as the wide dynamic range above the input signal of these radar coverage has been challenging. In this paper, a method for digital design and determination of the hardware architecture of a high repetition frequency radar signal processor based on the use of software provided by Xilinx XSG It's easier to design and develop The FPGA-based chip-based is provided in the MATLAB Simulink software is presented, this method is based on the use of Software provided by Xilinx Inc. The results of hardware simulation and comparison of output blocks of processing blocks with the output of the analog blocks of the typical radar and comparison with analog digital combined hardware of general radars represent improvement dynamic range of input at least 70 dB and low weight of this processor for a radar with high pulse repetition frequency.https://jasp.tabrizu.ac.ir/article_9183_a46514d0a3fdb36bd82cd1e2b41b9aa9.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Electromyogram Signal Compression Based on Empirical-Mode-Decomposition-Based Approximation and DCT-Based SmoothingElectromyogram Signal Compression Based on Empirical-Mode-Decomposition-Based Approximation and DCT-Based Smoothing8396918410.22034/jasp.2019.9184FAM.MagariFaculty of Electrical and Robotics Engineering, Shahrood University of Technology, Shahrood, IranH.GrailuFaculty of Electrical and Robotics Engineering, Shahrood University of Technology, Shahrood, IranJournal Article20171116Electromyogram (EMG) signals are useful in muscle behavior assessment and have some clinical applications. Today, there is a great tendency to transmit and store long-term EMG recordings which implies the importance of EMG signal compression. In this paper, we have proposed an EMG signal compression approach based on Empirical-Mode-Decomposition-based signal approximation, Discrete-Cosine-Transform-based signal smoothing, two-dimensional signal processing, wavelet transform, and SPIHT coding. We have evaluated the compression performance of the proposed approach by two sets of measures: The compression throughput and clinical-information-preserving measures. The former include two measures of PRD and CF while the latter uses four spectral parameters as the appropriate measures.Electromyogram (EMG) signals are useful in muscle behavior assessment and have some clinical applications. Today, there is a great tendency to transmit and store long-term EMG recordings which implies the importance of EMG signal compression. In this paper, we have proposed an EMG signal compression approach based on Empirical-Mode-Decomposition-based signal approximation, Discrete-Cosine-Transform-based signal smoothing, two-dimensional signal processing, wavelet transform, and SPIHT coding. We have evaluated the compression performance of the proposed approach by two sets of measures: The compression throughput and clinical-information-preserving measures. The former include two measures of PRD and CF while the latter uses four spectral parameters as the appropriate measures.https://jasp.tabrizu.ac.ir/article_9184_12c35c5392959177781ba38936ae7450.pdfVice Chancellery for Research and Technology, University of TabrizAdvanced Signal Processing2676-33973120190522Farsi Speech Synthesis Using Pitch Frequency in Flite SoftwareFarsi Speech Synthesis Using Pitch Frequency in Flite Software97107918510.22034/jasp.2019.9185FAF.NaiemiElectronic Group, Semnan Branch, Islamic Azad University, Semnan, IranV.GhodsYoung Researchers and Elite Club, Semnan Branch, Islamic Azad University, Semnan, IranJournal Article20170806This survey introduces a model and the implementation of a speech synthesizer in Farsi language using Flite software. In this approach, the mean and the standard deviation of pitch frequency of each voiced phoneme are first calculated by a database of Farsi sentences (Fars Dat). Then, the changes of each phoneme of the desired phrase are inserted into the software through the calculation of a value. The main feature of this synthesizer is its ability to change text to speech within Farsi pronunciation and in Farsi dialect. At the end of this paper, the results of this algorithm are compared to the changes of pitch frequencies extracted from the database of Farsi sentences. Some examples of the sentences from the database are also synthesized using our proposed method on Flite Software. The value of MOS test for understandability, naturalness and good sounding of those sentences are 4.4, 4.2, and 4.6 for the training set, respectively, and 4.2, 4.1, and 4.3 for the test set, respectively.This survey introduces a model and the implementation of a speech synthesizer in Farsi language using Flite software. In this approach, the mean and the standard deviation of pitch frequency of each voiced phoneme are first calculated by a database of Farsi sentences (Fars Dat). Then, the changes of each phoneme of the desired phrase are inserted into the software through the calculation of a value. The main feature of this synthesizer is its ability to change text to speech within Farsi pronunciation and in Farsi dialect. At the end of this paper, the results of this algorithm are compared to the changes of pitch frequencies extracted from the database of Farsi sentences. Some examples of the sentences from the database are also synthesized using our proposed method on Flite Software. The value of MOS test for understandability, naturalness and good sounding of those sentences are 4.4, 4.2, and 4.6 for the training set, respectively, and 4.2, 4.1, and 4.3 for the test set, respectively.https://jasp.tabrizu.ac.ir/article_9185_08e728bbe0f267dee5dd6f9c4245a152.pdf