profile - Razi University

Faculty Member of Razi University

Razi University
Mohammad Saeed Jahangiry

Mohammad Saeed Jahangiry

Assistant Professor / Engineering / Dept. of Computer Engineering

Current courses

Course Name unit term
Digital Circuits 3 first semester Academic year 2026-2027
Interface Circuit Design 3 first semester Academic year 2026-2027
3 3 first semester Academic year 2026-2027

Master Theses

  1. ارائه يك چارچوب يادگيري چند وظيفه اي دو شاخه مبتني بر شبكه هاي عصبي عميق بر پايه Resnet18 و FaceNet جهت آشكارسازي، تطبيق چهره و تشخيص حملات جعل هويت
    Samira Salimi ghaleh 2026
       Abstract With the increasing deployment of face recognition and verification systems in biometric applications, securing these systems against presentation attacks (spoofing) has become a critical challenge in real-world implementations. Attacks such as the use of printed images or video replays can severely compromise the performance of face-based authentication systems. Therefore, designing a framework capable of performing both face verification and presentation attack detection (PAD) simultaneously is of significant importance. In this research, a multi-task learning (MTL) framework based on a dual-head architecture is proposed to perform both tasks concurrently. In this architecture, a shared backbone is employed for feature extraction, followed by two distinct heads tailored to each task. To evaluate this framework, two deep networks, ResNet-18 and FaceNet, were utilized and compared within the proposed ResNet-DualHead and FaceNet-DualHead configurations. Evaluations on the standard LCC-FASD dataset demonstrate that the ResNet-DualHead architecture achieves superior and more stable performance in face verification, with an accuracy of 93.00%, an AUC-ROC of 97.87%, and an EER of 7.93%, compared to FaceNet-DualHead (88.00% accuracy, 93.89% AUC-ROC, and 12.56% EER). Furthermore, in the task of presentation attack detection, the ResNet-DualHead architecture achieves an APCER of 5.33%, a BPCER of 8.53%, and an ACER of 6.93%, effectively identifying a significant portion of spoofing attacks compared to the FaceNet-DualHead (APCER 8.75%, BPCER 14.89%, and ACER 11.82%). In conclusion, the results of this study indicate that the dual-head architecture, implemented within a multi-task learning framework and utilizing ResNet-18 as a shared backbone, provides an effective and lightweight solution for mitigating presentation attacks in face recognition systems, enabling real-time deployment on resource-constrained edge devices. Keywords: Face Recognition, Face Verification, Presentation Attack Detection (Spoofing), Multi?Task Learning, Dual?Head Architecture, ResNet-18, FaceNet
  2. Design and construction of a two-axis light and sun tracking system using Arduino and LDR sensors
    Mohammad Javaad Rajabi 2026
      طراحي و ساخت سيستم دنبال كننده خورشيدي دو محور با استفاده از آردوينو و حسگر LDR
  3. Electromyography-based hand gesture recognition using machine learning methods
    Zahra Jahani Mehrabadi 2026
    Physiological signals provide various types of information about the human body. The diagnosis, monitoring, and treatment of diseases are made possible through the accurate acquisition of these signals, which are obtained using either invasive or non-invasive sensors. Sensors or electrodes placed on muscles provide data regarding electrical muscle activity, known as Electromyography (EMG). Gestures involving body parts, such as the face and hands, facilitate communication with others. Coordinated muscle activity produces diverse hand movements, which can be learned and predicted using machine learning algorithms following a training phase. Electromyography (EMG) pattern recognition has long been studied for gesture recognition and the control of prosthetics and rehabilitation systems. In recent years, hand gesture recognition based on EMG signals has emerged as a prominent field within human-computer interaction, assistive technologies, and the control of artificial limbs. The objective of this thesis is to design, implement, and evaluate a complete machine learning pipeline for >In this study, following preprocessing (training/test splitting and normalization via StandardScaler) and a multifaceted exploratory signal analysis—encompassing time-domain examination, dimensionality reduction (PCA and t-SNE), RMS calculation, inter-channel correlation analysis, and frequency analysis (Power Spectral Density and spectrograms)—two >A dataset was utilized to technically validate this pipeline at the current stage of the research. Both the Random Forest and CNN models achieved 100% >  
  4. مديريت تخصيص منابع محاسبات چند مه در وسايل نقليه خودران
    Mohammadhadi Akbarzadeh 2025
  5. Distributed event-triggered control in DC microgrid under cyber attacks
    Omid Danaei Koik 2025
    Direct current (DC) microgrids have recently attracted more attention from researchers due to their advantages over alternating current (AC) microgrids, such as the absence of transmission losses related to reactive power flow, the absence of harmonic currents, and the simple integration of resources with DC loads. However, the constant exchange of information in the secondary control layer in the cyber environment for voltage recovery and optimal resource management has presented these microgrids with challenges such as high communication costs, the need for high communication bandwidth, and cyber threats. Denial of service (DoS) attacks and false data injection (FDI) attacks are dangerous examples of these threats that can lead the system to instability and even collapse. DoS attacks have extremely destructive effects on the system by temporarily or permanently disrupting access to critical system information. On the other hand, FDI attacks can secretly and noticeably cause harmful disturbances in the system by injecting false data into the transmitted information. In this study, an edge-based event-triggered control structure is proposed as a suitable and optimal alternative for permanent information exchange in second-layer distributed control, which is able to ignore DoS attacks. Based on this structure, information exchange between units occurs only when the sending conditions are met. Also, a decentralized approach based on a multilayer perceptron (MLP) neural network is introduced to detect FDI attacks. The advantages of this approach include high flexibility, reduced information transmission in the cyber environment, reduced complex calculations, and increased information security. Also, a threshold for an FDI attack is determined so that mitigation operations are performed only when the attack occurs. A microgrid consisting of three generating units and a DC voltage bus is simulated under different scenarios to investigate the effectiveness of this study. The results obtained show that the event-triggered (ET) structure is resistant to DoS attacks and the proposed approach is able to detect FDI attacks with high accuracy in the presence of DoS attacks and in the ET communication structure and mitigate its effects. Keywords: Direct Current Microgrids, Denial of Service Attacks, False Data Injection Attacks, Edge-based Event-triggered, Distributed Control, Multilayer Perceptron Neural network
  6. Voltage estimation in DC microgrid using neural network
    Farshid Dadsetan 2024
       As we know, the structure of electricity networks around the world is changing and evolving. So that distributed production in the form of micro-grid (MG) is expanding. Normally, an MG consists of several scattered production sources, and energy storage systems (batteries) are used for continuity of energy production in new MGs. In such networks, determining the charging and discharging time of the batteries requires having sufficient information about the state of energy production and consumption or the size of the MG bus voltage. Normally, sensors are used in MGs to get information about the above-mentioned things. Using a sensor in any system reduces its reliability. Because there is a possibility of failure in the sensors. For this purpose, the use of sensorless methods to estimate the MG bus voltage creates better conditions in creating more confidence in the MG's reliable performance. Therefore, in this research, an artificial neural network (ANN) model is proposed for MG modeling. Because A   have shown to be capable of modeling non-linear systems. It should be noted that the MG under study in this work consists of a distributed production unit (solar energy) and an energy storage system unit (batteries). In this way, three ANN models have been created to estimate MG bus voltage. So that the first model estimates the output voltage of the photovoltaic array, the second model estimates the current of the storage system, and the third model estimates the MG bus voltage. To evaluate the performance of the created models, the mean square error (MSE) of each model has been calculated. The test results show that the error of the proposed models is very low and close to zero, which means that the models can predict the output with high accuracy. Finally, it should be said that the accuracy of the first model that estimates the voltage of the photovoltaic array is higher than the other two models.
  7. Disease epidemic modeling based on population structures to optimize the allocation of medical resources
    Niloufar Jafari 2023
    با توجه به شيوع گسترده بيماري هاي واگيردار در جهان مدل‌هاي رياضي مي‌توانند به پيش‌بيني و كنترل اين پاندمي كمك كنند. پاندمي بيماري نشان داده كه كشورها عليرغم توسعه يافتگي و دسترسي به منابع و تجهيزات در مقابل يك بيماري جديد ناشناخته بايستي زير ساخت‌هاي فيزيكي اعم از بيمارستان‌ها، مراكز بهداشتي درماني و نيروي انساني را در برابر بحران آماده نمايند. اپيدمي ها ممكن است ظرفيت بيمارستاني ارائه خدمات بهداشتي درماني را در هم بشكنند و منابع مادي و انساني، از جمله فضاي بيمارستان، تجهيزات و داروها براي برآوردن تقاضا كافي نباشند، مخصوصاً در مورد اپيدمي كه چندين هفته يا چندين ماه طول بكشد و به ويژه اگر بحران‌هاي همزمان اتفاق افتد، بيمارستان جهت كمك به اقدامات انجام شده در كنترل اپيدمي، بايد بسياري از كاركردها و منابعش را مهار كرده و به طور هماهنگ شده مورد استفاده قرار دهدكه برآوردن اين الزامات مي‌تواند چالش برانگيز باشد. يك اپيدمي نياز به مركز بهداشتي درماني دارد تا اولويت‌هايش را تغيير داده و با روش‌هاي كاري منطبق باشد تا پاسخي سيستميك و هماهنگ به يك موقعيت با   سرعت در حال تغيير بدهد.   بر اساس پيش‌بيني و صحت نتايج مربوطه، آماده‌سازي مراكز درماني در مقابل بحران از جنبه‌هاي مختلف مديريتي، ارتباطات، منابع انساني، تامين تجهيزات، دارو و ساير خدمات تشخيصي، درماني و پشتيباني مورد نياز بيماران، همراهان و كاركنان بهداشت و درمان مهم‌ترين هدف اين تحقيق مي‌باشد كه با مدل‌سازي همه‌گيري بيماري در شرايط مختلف محقق مي‌شود. شبيه سازي با در نظر گرفتن مواردي مانند: تاثيرات اندازه جمعيت، دوره نهفتگي بيماري، دوره بيماري، نرخ آلودگي، مكان‌هاي آلوده بر گسترش بيماري، واكسيناسيون عمومي و ... اجرا و تحليل خواهدشد.
  8. Adjustable antenna design with resetting at the radiation angle
    Milad Mohamadkhani 2023
    The design of a reconfigurable antenna that is able to change its radiation pattern characteristics at a certain frequency in an organized and reversible manner using control switches such as (semiconductor switches, microelectromechanical switches and varactor switches), and application in the fifth generation of wireless communication networks have The present research method is descriptive-analytical and practical in terms of applying the results. Collecting the required data and information is also done by reading articles, books, using the knowledge of professors and observing the microstrip antennas that have been designed and built in the past, and to analyze and analyze the obtained data and simulate the desired models, high frequency circuit simulation software is used. CST has been used. In the research, a microstrip antenna (printed circuit patch antenna) has been designed to achieve reconfiguration of the radiation pattern. Two capacitors work simultaneously to achieve the radiation pattern reconfiguration operation. This microstrip antenna is i  ired by the design of a circular loop. The basic structure of the loop is changed and the capacitors are integrated into the patch. By changing the capacity of the capacitors in the range of (0.01-10) picofarad, the radiation pattern of the antenna can be changed to six different modes. An electromagnetic model of the proposed antenna is simulated in CST software for numerical analysis and observation of different radiation patterns. The proposed antenna structure has been implemented and built using a material (FR4) substrate with relative transmittance (4.4) and loss tangent (0.02) and thickness (1.6) mm. Antennas have been used since 1886 to send and receive information. They have special characteristics, this divides antennas into two categories, the first category is antennas with fixed characteristics and the second category is tunable antennas, tunable antennas are divided into several categories, the main ones are tunable antennas (frequency, polarization, bandwidth and radiation pattern), these types of antennas are very useful in the technology of the fifth generation of wireless communication systems. One of the achievements of this research is the design and simulation of a microstrip antenna that can reset the radiation pattern by using capacitors as control switches in order to change the surface current and as a result change the radiation pattern of the antenna.   
  9. Applying data mining to forecast generated power by solar power plant used in probabilistic optimal power flow
    Negin Fatahnia 2022
    Abstract  The increasing influence of renewable energy sources, including solar energy, in the production of electrical energy, has increased uncertainties in solving various problems,control and operation of the power system.The increasing influence of renewable energy sources, including solar energy, in the production of electrical energy, has increased uncertainties in solving various problems, control and operation of the power system.The increasing influence of renewable energy sources,including solar energy, in the production of electrical energy, has caused an increase in uncertainties in solving various problems, control and exploitation of the power system.this reason, probabilistic load distribution has become an important tool for investigating the random characteristics of the power system.In this thesis, in order to investigate the amount of losses, considering the unstable behavior of loads in the form of a normal distribution, one of the Monte Carlo probabilistic load distribution methods is used.Two-point, three-point and five-point estimation are used with the help of Matlab software. In the following, the MCS&3PEM algorithm is proposed and the results obtained from it are compared with the Monte Carlo method.Also, in order to check the amount of fuel cost, the PSO algorithm has been used; PSO The obtained results will be evaluated to find the best location of photovoltaic systems, in order to reduce the cost of fuel. Two standard test systems are used to check the said methods. Also, to estimate the amount of solar radiation received in Kermanshah province using daily data obtained from meteorological station, in a period of 6 years, GMDH optimized neural network is proposed. Also, to estimate the amount of solar radiation received in Kermanshah province using daily data obtained from meteorological station, in a period of 6 years, GMDH optimized neural network is proposed.Also, to estimate the amount of solar radiation received in Kermanshah province using the daily data obtained from the weather station, in a period of 6 years, an optimized GMDH neural network is proposed.The results of the research showed that there is a small difference between the radiation values measured at the meteorological station and the radiation obtained by the model, which shows the ability of the model to estimate the radiation. The results of the research showed that there is a small difference between the radiation values measured at the meteorological station and the radiation obtained by the model, which shows the ability of the model to estimate the radiation.The results of the research showed that there is a small difference between the radiation values measured at the meteorological station and the radiation obtained by the model, which indicates the ability of the model to estimate radiation.In order to evaluate different modes of system operation, two scenarios areconsidered. First, all the loads are fed only by the main grid, and losses and production costs are expected to be high. And second, the photovoltaic system is connected to the system. Based on the results,In order to evaluate different modes of system operation, two scenarios are considered. First, all the loads are fed only by the main grid, and losses and production costs are expected to be high. And second, the photovoltaic system is connected to the system. Based on the result،In order to evaluate different modes of systemoperation, two scenarios are considered. First, all the loads are fed only by the main grid and the losses and production costs are expected to be high. And second, the photovoltaic system is connected to the system. Based on the It was found that when the photovoltaic system is connected to the system, the amount of power loss and production cost   is reducedKeywords:
  10. Protection of microgrids using voltage-based power differential and sensitivity analysis
    Ahmad Mirzaei 2022
    امروزه با ورود منابع تجديدپذير به شبكه توزيع موضوعي به نام ريز شبكه شكلگرفته كه اين موضوع بخشي ازتحقيقات را به خود تخصيص داده است. ريزشبكه داراي مزيتهاي بسيار زيادي ازجمله طبيعي و اقتصادي و غيره است. اگرچه ريزشبكه داراي مزاياي بسيار زيادي است، اما ريز شبكه چالشهاي متفاوتي را نيز ايجاد كرده است. يكي از چالشهاي مهم ريزشبكه موضوع حفاظت از ريزشبكه است، كه موضوع موردبحث اين پاياننامه نيز هست. چون ريزشبكه در دو مد كاري ايزوله و متصل به شبكه كار ميكند، به همين علت بر اندازه جريان اتصال كوتاه به شدت تأثير ميگذارد. در مد كاري متصل به شبكه اندازه جريان اتصال كوتاه متشكل از جمع جريانهاي اتصال كوتاه توسط شبكه اصلي و DG هاي موجود در ريزشبكه ميباشد، اين در حالي است كه در حالت ايزوله مقدار جريان اتصال كوتاه فقط توسط DG هاي موجود در ريزشبكه تغذيه ميشود. در نتيجه به علت تغييرات سطح جريان اتصال كوتاه و وابستگي اندازه اين جريان به توپولوژي شبكه، سيستم حفاظتي اضافه جريان، توانايي حفاظت از ريزشبكه را ندارد. به همين سبب در اين پاياننامه راهكاري براي حفاظت از ريزشبكه ارائه ميشود . الگوريتم بررسيشده در اين پاياننامه از طريق محاسبات حساسيت و تغييرات توان اكتيو بر پايه اندازهگيري فازور ولتاژ به صورت همزمان در يك زون حفاظتي، خطا را شناسايي كرده و فرمان تريپ را صادر ميكند. براي بررسي بيشتر اين الگوريتم از شبكه استاندارد CIGRE در برنامه ديگسايلنت استفاده شده و الگوريتم پيشنهادي با زبان برنامه نويسي پايتون در برنامه ديگسايلنت در قسمت DSL شبيهسازيشده است. مزيت اين الگوريتم براي حفاظت از ريزشبكه اين است كه، اولا توپولوژي شبكه بر كاركرد اين الگوريتم تاثير ندارد و ثانين توانايي تشخيص انواع خطا در دو مد كاري ريزشبكه را نيز داد. در پايان براي كاركرد بهتر اين الگوريتم به قسمت تصميمگيري آن يك شرط براي ولتاژ اضافه شده، كه باعث برطرف شدن تصميمگيري اشتباه توسط اين الگوريتم در زمان خطاهاي طولاني مدت ميشود
  11. The Investigation of changes in the shear stress pattern in a rectangular channel in terms of changing geometric characteristics and provide design solutions.
    MOHAMMAD JAVAD KARIMI 2022
  12. Studying the effects adding silica gel fiber on the mechanical properties of Roller compacted concrete pavement
    Hamed Ardalan 2021
  13. Evaluation of the effect of formal education on people's awareness of traffic regulations and traffic safety
    Shahryar Moradi 2021
  14. Numerical study of creating step in soil nail walls with considering of corner effect
    Yaser Ahmadbeigi 2021
       پايدارسازي گود و شيب ها در مناطق مختلف به دليل گسترش جمعيت ،رشد شهرسازي و راهسازي جهت تامين امنيت جاني و جلوگيري از خسارات مالي   امروزه مورد توجه بسياري از محققين قرار گرفته است . يكي از رايج ترين روشهاي پايدار سازي به دليل مزاياي بسيار ديوارهاي ميخ كوبي شده است. اخيرا برم بندي در ديوارهاي ميخ كوبي مورد پژوهش قرارگرفته و نتايج مثبت آن به وسيله مطالعات دوبعدي نشان داده شده است. باتوجه به هندسه هاي مختلف و مؤثر بر رفتار ديوار نظير گوشه ها و قوس ها با انجام مطالعات سه بعدي ميتوان رفتار ديوارهاي ميخ كوبي شده در اين شرايط را بهتر شناخت و منجر به ديد مفيد تري از عملكرد آنها شد . دراين پژوهش با استفاده از نرم افزار پلكسيس سه بعدي و دوبعدي اثرات ديوارهاي ميخ كوبي برم بندي شده را در حالت همزماني با گوشه در گود ها ، در يك تحليل عددي مورد بررسي قرار گرفته و نتايج مثبت همزمان ناشي از وجود گوشه و برم بندي وهمچنين تغييرات عرض آن بر روي نشست ها(از مقدار 9 سانتيمتر نشست به 1 سانتيمتر در عرض سه متر كاهش يافته)، تغيير مكان هاي افقي ديوار و پايداري بررسي   قرارگرفته و مدل رفتاري بررسي و ضعف مدل موركولمب نشان داده شد .
  15. Design of ultra wide band low noise amplifier with noises canceling technique with capacitive feeder network
    Hossein Khodarahmi 2021

Update: 2026-11-18

افراد توان خواه
رنگ متن
رنگ پس زمینه
اندازه فونت
+ A -
پنهان سازی تصاویر
تنظیم کنتراست
فاصله بین کلمات
خط راهنما
فاصله بین خطوط
نوع فونت
بزرگنمایی انداره صفحه
+ -
توقف انیمیشن ها