Thesis published in Dept. of E.E.E http://103.99.128.19:8080/xmlui/handle/123456789/33 2026-04-09T19:30:12Z 2026-04-09T19:30:12Z Das, Diba http://103.99.128.19:8080/xmlui/handle/123456789/511 2025-09-23T05:25:44Z 2023-10-15T00:00:00Z

Title: Hardware-Software Codesign of Electrooculogram Signal Processors Authors: Das, Diba Abstract: Electrooculogram (EOG) is a bioelectric signal carrying eye movement information. This signal can be utilized in medical and bio-electrical applications such as diagnosing different ocular diseases and controlling human-computer interfaces. Point of care (POC) systems refer to the systems where testing is performed right where the patient is. POC systems dedicated to detecting various eye conditions can be developed by hardware implementations of EOG. Field Programmable Gate Arrays (FPGAs) are reconfigurable integrated circuits offering flexible software features with fast parallel computation. Application specific designs in FPGA fix the functionality and reduce the number of components used providing cost-effectiveness, and power efficiency. In this research, a systematic investigation of the research trend on hardware implementations of EOG is presented first. After that, an in-depth analysis of two novel FPGA-based architectures is presented. The first work aims to design a hardware-optimized binary EOG processor for blink detection by using multichannel EOG signals containing horizontal and vertical EOG signals. After preprocessing the EOG signals, by extracting only two features- root mean square (RMS) and standard deviation (STD), blink and saccades are classified employing support vector machine (SVM) with 97.5% accuracy. The implemented system of this design in Xilinx Zynq-7000 FPGA achieves an accuracy of 95%. The second work aims to design a hardware-optimized machine learning based EOG signal processor to classify six different eye movements (up, down, normal, right, left, and blink) adopting an SVM classifier with average software accuracy of 97.92%. The accuracy of the implemented system in Zynq UltraScale+ is 95.56%. The efficacy of the developed systems is proved by comparing them with state-of-the-art technologies. A few potential future research scopes are mentioned at the end of the thesis. Description: An M.Sc. Thesis from the Department of Electrical and Electronic Engineering

2023-10-15T00:00:00Z Tasnim, Moshammed Nishat http://103.99.128.19:8080/xmlui/handle/123456789/510 2025-09-23T05:25:32Z 2024-03-28T00:00:00Z

Title: Design and optimal power management technique for resilient operation in an interconnected hybrid microgrid system Authors: Tasnim, Moshammed Nishat Abstract: The concept of networking multiple microgrids is one of the most promising initiatives in microgrid-based power generation frameworks to address the challenges of a single microgrid's resiliency and enhance supply security. Interconnected microgrids have the potential to serve as a fundamental framework for future distribution systems due to the extensive deployment of smart grid technology. Infrastructure planning and design, control theory, and communication technologies are required to regulate microgrid clusters in a flexible and efficient manner. That is, a proper control structure with robust and reliable control strategies is required to improve the performance of the interconnected microgrids in terms of power-sharing, power quality, and stability. The study introduces a novel control structure and a clustering method for interconnected hybrid microgrids to form an integrated framework. The proposed control structure for power flow control among hybrid microgrids places particular emphasis on the control strategies of three converters, such as the energy storage system, the interlinking converter of each hybrid microgrid, and the interconnecting converter for networking multiple microgrids. The control strategy of the hybrid microgrid’s interlinking converter is based on a voltage-frequency droop control to ensure proper operation in three operating modes, such as islanded, grid-connected, and interconnecting modes. The virtual inertia and state-of-charge-based controller of the energy storage system controls the battery bank's charging and discharging operations and provides an autonomous power flow inside each hybrid microgrid. Finally, the control strategy of a parallel interlinking converter structure is designed to interconnect and control the power flow among hybrid microgrids. vi The control structure allows both islanded and grid-connected operations without swapping between two controllers. This control framework reduces the number of activation operations in response to variations in conditions. Consequently, negative consequences like an uneven transition and system failure due to an inaccurate transition can be mitigated. In summary, the suggested control framework can function seamlessly upon the occurrence of unintended situational changes, such as main grid failure or islanding (with -0.11% variation in frequency), load variation (-0.038% variation during load increment and 0.02% variation during load decrement), source failure (-0.13% and 0.18% variations in frequency during wind generator failure), and grid synchronization (with 0.04% variation in frequency), without requiring a control mode transition. The clustering method and control structure of the interconnected microgrid system are designed in a MATLAB/Simulink environment. The OPAL-RT simulator (OP5600)-based real-time software-in-the-loop simulation technique is used to analyse the performance of the interconnected system in terms of load variations, source failure, power quality, mode transition, and energy storage system management. The results indicate that the control structure, with three control strategies, ensures reliable operation in all modes with reduced total harmonic distortion (less than 5%) and lower frequency fluctuation (less than 1%), and also maximizes power supply security. Description: EEE thesis for Master of science

2024-03-28T00:00:00Z Khondaker, Md. Fazlul Karim http://103.99.128.19:8080/xmlui/handle/123456789/507 2025-09-23T05:24:03Z 2024-03-05T00:00:00Z

Title: EEG-based preference prediction in neuromarketing using objective labeling and investigating the effects of languages on the preferences of consumers Authors: Khondaker, Md. Fazlul Karim Abstract: Neuromarketing is an emerging brain-computer interface (BCI) research field that aims to understand consumers' internal decision-making processes when choosing which products to buy. It provides valuable insights for marketers to improve their marketing strategies based on consumers' impressions. However, the current status of Electroencephalography (EEG)-based preference prediction and its classification accuracy is still below optimal. The performance of EEG-based preference detection systems depends on a suitable pre-processing pipeline selection and proper data labelling since noisy EEG data and wrong labeling are likely not to give better results. Most of the previous studies followed a traditional EEG pre-processing pipeline, and also recently, an advanced automated EEG pre-processing pipeline has been introduced in the literature. In this study, I have proposed an optimal EEG pre-processing pipeline and compared it with the other two existing methods. I collected raw EEG data and pre-processed them using the three pre-processing pipelines. I then extracted many statistical and frequency domain features and employed several machine learning (ML) classification models. After implementing my proposed pipeline to pre-process EEG data, I observed a significant improvement in classification accuracies compared to the other two existing methods. Another essential thing is to label the data properly so that ML models can be trained accurately. Previous studies have used different ML models to predict subjects' future preferences based on subjective labeling, i.e., the preferences are self-reported by the subjects. However, such labeling methods contradict the main goal of Neuromarketing research, which is to detect genuine preferences from brain data rather than rely on subjects-reported data. This leads to the necessity of an objective labeling approach where the labels can be detected directly from the brain data. In this study, I proposed an objective labeling method for EEG-based preference prediction in Neuromarketing. I compared the performance of different ML models trained on objective and subjective labeling using two datasets: a publicly available Neuromarketing dataset and a new dataset created in my experiments. My findings demonstrate that the ML models trained on objective labeling provided better classification results than the ones trained on subjective labeling. The result aligns with the goals of Neuromarketing research, as it offers an automated way of labeling data and opens up new avenues for future research. Also, among the prominent components of promotions, language has a major impact on consumers' minds. I investigated the effects of foreign languages (FL) on consumers' preferences in Neuromarketing and found that consumers show risk adoption tendencies when exposed to product ads in FL instead of their native languages (NL). Description: EEE

2024-03-05T00:00:00Z Chowdhury, Aditta http://103.99.128.19:8080/xmlui/handle/123456789/504 2025-09-23T05:20:27Z 2023-10-01T00:00:00Z

Title: Digital Design of Photoplethysmogram (PPG) based Cardiovascular Disease Classifiers Authors: Chowdhury, Aditta Abstract: Photoplethysmogram is an optically obtained signal working based on the volumetric change of blood. As heart diseases are correlated with the pumping of blood, PPG can be studied for detecting cardiovascular diseases. Researchers have already analyzed PPG signals for various disease detection, including hypertension, coronary artery disease, diabetes, and others. Also, two important health parameters: heart rate and blood pressure have been predicted from PPG signals in several studies. However, most of the work has been done at the software level without any hardware implementation. In addition, two important cardiovascular diseases related to blood flow in the brain: cerebral infarction and cerebrovascular disease are yet to be explored based on PPG signal. Hence, this study aims to develop a hardware-based system that can detect several cardiovascular diseases - hypertension, cerebral infarction, cerebrovascular disease, diabetes, and a few combinations of them. The study checks the feasibility of detecting these diseases individually in a binary classification system and also in a multiclass classification system. A system is also implemented for predicting heart rate and blood pressure from PPG signals. The systems are developed in Xilinx system generator targeting Zedboard zynq 7000 and zynq ultrascale+ FPGA board. The binary classification system uses 11 features and applied SVM classifier to get the accuracy of 96.37%, 93.48%, 96.43%, and 88.46% for detecting hypertension, cerebral infarction, cerebrovascular disease, and diabetes, respectively, consuming a total of 0.693 W power. The multi-class classification system utilizes a total of 1.403 W of power, providing an accuracy of 79.83% for detecting 7 classes of diseases. Also, the heart rate and blood pressure estimation system utilizes 0.353 W of power. The heart rate is predicted with 4.04% error while systolic and diastolic blood pressure are estimated with 3.77% and 4.8% error, respectively. The designed prototype can be further extended to develop wearable devices, and smartwatches and can be useful for medical treatment and analysis. Description: An M.Sc. Thesis from the Department of Electrical and Electronic Engineering

2023-10-01T00:00:00Z