Scholarly Indexed Publications

Permanent URI for this collectionhttps://hdl.handle.net/10361/28605

Browse

Now showing 1 - 20 of 511

Metadata Only
Real-time object detection and distance measurement for the patients with visual-spatial agnosia
(Institute of Electrical and Electronics Engineers Inc., 2024-01-01) Das, Adri Shankar; Sahill, Fardin Rahman; Fahad, Ababil Hossain; Shakib, Mohammed; Rahim, A.H.M.A.; Mahmud, Tasfin; Shawon, Md. Mehedi Hasan
Visual-spatial agnosia is a neurological visual impairment disorder that occurs after brain strokes in elderly people. These patients are unable to recognise objects surrounding them and also they cannot pursue the distance of the object. This creates a huge barrier between the real world and the world they perceive. This study proposes a smart eyewear assistive device that will help patients recognize the object as well as guide the patient toward the object by showing the distance of the object with the help of machine learning algorithms. A stereo vision algorithm is used to estimate the distance along with the MaskRCNN algorithm that detects the objects. Physical data has been recorded to analyze the models' performance with the MaskRCNN algorithm with stereo vision algorithm. This study also proposes future works in this area.
Metadata Only
Implementation of hector SLAM algorithm for mapping indoor environments with obstacles
(Institute of Electrical and Electronics Engineers Inc., 2024-01-01) Hossain, Amreen; Chowdhury R.H.
Autonomous navigation in obstacle-rich indoor environments is crucial for both industrial and domestic robotic applications. A central aspect of this process is Simultaneous Localization and Mapping (SLAM). This paper presents a comprehensive methodology for implementing Hector SLAM, a widely used SLAM algorithm that operates without relying on external odometry and pose data from wheel encoders or Inertial Measurement Unit (IMU) sensors. The experimental setup involves a custom-built two-wheeled mobile robotic platform, equipped with essential components including a 2D Light Detection and Ranging (LiDAR) sensor, wheel encoders, IMU, motor drivers, an Arduino UNO, and a Raspberry Pi 4B, with Robot Operating System (ROS) serving as the primary software framework. The robot navigates within a predefined rectangular indoor environment, where real-time mapping results are captured at different time intervals to assess mapping accuracy. Additionally, a randomly placed object is introduced to test the obstacle detection capability during operation. Results indicate that Hector SLAM demonstrates high accuracy, achieving precision within a few centimeters, particularly in environments with distinct structural features.
Metadata Only
Enhancing trajectory tracking of Quadrotor using feedback linearization with MPC-LPV and LQI-LPV under variable disturbances
(Institute of Electrical and Electronics Engineers Inc., 2024-01-01) Chowdhury R.H.; Imtiaz Ferdous A.; Hossain, Amreen; Rahman K.A.
The deployment of Unmanned Aerial Vehicles (UAVs) has markedly improved industrial efficiency. Optimizing control algorithms for precise path tracking is essential for enhancing the reliability and performance of these systems. This paper presents a comparative analysis of Model Predictive Controller (MPC) and Linear Quadratic Integral (LQI) controller, combined with the Linear Parameter Varying (LPV) approach, for quadrotor trajectory tracking under variable disturbances. The quadrotor is modeled using Newton-Euler equations for a six-degrees-of-freedom system. The control architecture includes an outer loop position controller using Feedback Linearization (FL) and an inner loop attitude controller implemented as either MPC-LPV or LQI-LPV. Disturbances are modeled using a custom sinusoidal model. Numerical simulations are done to evaluate controller performance for six different trajectories. The results for the helical trajectory indicate that under variable disturbances, the MPC-LPV+FL outperforms the LQI-LPV+FL. The MPC-LPV+FL achieves a steady-state error of 0.1934 m compared to 0.6611 m for the LQI-LPV+FL. Additionally, the MPC-LPV+FL reaches the minimum error in 3.1 seconds, compared to 3.8 seconds for the LQI-LPV+FL, resulting in a 22.58% faster response. Therefore, MPC-LPV+FL is the superior choice based on accuracy, stability, and speed of response under adverse conditions.
Metadata Only
Automated online exam proctoring system using computer vision and hybrid ML classifier
(Institute of Electrical and Electronics Engineers Inc., 2021-01-01) Hossain, Zarin Tahia; Roy, Protyasha; Nasir, Rina; Nawsheen, Sumaiya; Hossain, Muhammad Iqbal
Importance of online education can be seen especially during the ongoing Covid-19 when going to schools or colleges is not possible. So validity of online exams should also be maintained with respect to traditional pen-paper examinations. However, absence of invigilator makes it easy for the examinees to cheat during the exam. Though there are already many systems for online proctoring, not all educational institutes can afford them as the systems are very expensive. In this paper, we have used eye gaze and head pose estimation as the main features to design our online proctoring system. Therefore, the purpose of this paper is to use these features to create an online proctoring system using computer vision and machine learning and stop cheating attempts in exams.
Metadata Only
IoT based health monitoring using smart devices for medical emergency services
(Institute of Electrical and Electronics Engineers Inc., 2019-11-01) Shabnam, Farzana; Azmi Hoque, S.M.; Faiyad, Shahed Al
Internet of Things points to expanding network of physical objects. Here, the data accumulation is done by the usage of sophisticated sensors, which communicates with each other and stores the data in cloud. IoT is transforming the way people communicate, work and live. It has been integrated in our everyday life starting from smart home devices to medical industries. In this paper, we have aimed to provide a review of different types of IoT based health monitoring devices proposed by researchers and how they are monitoring specific diseases. Such devices are categorized based on their types and comparison has been made based on available features. The following paper will focus on different wearable health devices; their operations, limitations and challenges. Moreover, an application on how these devices could be used for emergency situations has also been discussed. It will help future researchers to have an overview on the recent advancements in health care monitoring system.
Metadata Only
Which matters more: Model or language? an empirical study in English-Bangla mental health classification
(Institute of Electrical and Electronics Engineers Inc., 2025-01-01) Islam, Apu; Rafi, Ishraque Arefin; Mondal, Sudipta; Rahman, Somaya Al Sadia; Alam, Golam Rabiul
We present an empirical comparison of classical baselines and pretrained transformer encoders for mental health status classification across English and Bangla social media text. Using two public datasets-an English multi-class corpus and a Bangla binary corpus-we evaluate TF-IDF with Logistic Regression and Random Forest against BERT, RoBERTa, DeBERTa, and BanglaBERT under a matched setup with a stratified eighty twenty split and macro F1 for model selection. In English, RoBERTa achieves 81.6% accuracy with a macro F1 of 78.8, while TF-IDF with Logistic Regression reaches 77.3% accuracy. In Bangla, BanglaBERT attains 88.3% accuracy with a macro F 1 of 88.3, and classical baselines surpass several non-Bangla encoders. Findings highlight the value of language models and the importance of classical machine learning models in classifying mental status across different languages.
Metadata Only
Precision repairs: Achieving accuracy during repair missions at the international space station
(Institute of Electrical and Electronics Engineers Inc., 2023-01-01) Rafi, Suhail Haque; Ul Islam Chowdhury, Md Mubin; Rakib, Mazharul Islam; Chowdhury M.; Rhaman, Md Khalilur
This paper proposes an autonomous approach for repairing leakage at the International Space Station (ISS) using the NASA Astrobee robot with the 3rd Kibo Robot Programming Challenge (KRPC) guidelines provided by JAXA. We bring into play image processing and analysis from the Astrobee Nav Cam data, circle detection algorithms, AR tag detection techniques, autonomous navigation, and a Visual Projective Localization And Mapping (VPLAM) approach for irradiating lasers with excellent precision at the Astrobee Robot Simulator. In addition, we discuss the various capabilities of the Astrobee Robots onboard the ISS and how they can facilitate the delivery of such an autonomous mission in emergencies. Lastly, we walk through ways to deal with randomness and uncertainty while operating, ensuring robustness and accuracy. Ultimately, we scored a precision of up to 0.03 cm on the KRPC online simulator.
Metadata Only
A chatbot based auto-improving health care assistant using RoBERTa
(Institute of Electrical and Electronics Engineers Inc., 2023-01-01) Aoyon, Rifat Sarker; Hossain I.
Artificial intelligence, a revolutionary invention of the modern era, is being utilised in many sectors to serve humanity. Like many other sectors, this technology is providing facilities to the health sector as well. Health care is one of the most essential sectors for the human being, which motivated us to do research to make an AI-based platform for helping people by providing health care facilities, and the knowledge of RoBERTa was used here to train our AI model. We have designed this platform in a very simple way so that patients do not face trouble. It is a healthcare chatbot where a person can set and update reminders for taking medicine or anything related to healthcare by writing a natural language text command. It is capable of providing any kind of health-related information, predicting diseases, suggesting corresponding solutions, and so on. The implementation of NLP techniques enables this chatbot to understand human-level language, so a person can easily use the mentioned features by using natural language commands. A major feature of this chatbot is the auto-improving feature, which can play a significant role in developing the ability to improve the performance of this chatbot by continuously iterating with users. This kind of conversational interaction helps the conversational partner develop the amount of data. Though this chatbot has the capability to deliver an accurate response without having a huge amount of data, increasing the amount of data will enhance its ability to provide a faster response, and this auto-improving algorithm will assist in achieving this goal. Moreover, prescription reading ability is included in this chatbot, which makes it easier to set reminders for eating medicine.
Metadata Only
A RNN based self-learning audio generating chatbot for French language learners
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Aoyon, Rifat Sarker; Ara, Yamin; Baptee, Tahsin Anzum; Afroz, Mehrin; Reza, Md Tanzim
A chatbot is an artificial intelligence-based virtual conversational agent that is one of the most potential technologies in the modern era. There are so many sectors where chatbots are providing outstanding performance. This technology has the ability to be an assistant for those who are trying to learn a new language. That is why we have done research to make a chatbot based on an RNN algorithm in order to help people learn French. Although there are many online platforms available to learn French, many of these are not sufficient enough to learn a language as it is mandatory to have a conversational partner to practice a language. Without practicing a language with anyone, it is difficult for a person to keep memorizing a lot of vocabulary. French pronunciation rules are not the same as English. Therefore, we have added a feature for sending text along with audio messages and getting replies in French audio alongside text. Because of having these features, users will not only know the French vocabulary and grammatical rules but can also be introduced to French pronunciation systems and learn to speak French with a native French accent. The self-learning feature assists this chatbot to increase its capability of giving correct responses to the users day by day by having continuous conversations and removes the problem of collecting so much data to make this kind of chatbot. Also the paraphrasing feature will help users to get different replies to a particular question at different times, enriching their vocabulary and familiarizing them with the variety of replies to a particular question.
Metadata Only
A compact and lightweight astronaut assistance rover with robust validation
(Institute of Electrical and Electronics Engineers Inc., 2025-01-01) Aunjum, Ragib; Ahsan, Ali; Jafar, Zareef; Ahmed, S.M. Masrur; Issa, Razin Bin; Ali, Mahbub; Mahmud, K.M. Fahim; Islam, Mohammad Zahirul; Mohammad, Zaber; Johan, Gazi Musa Al; Islam, Saiful; Alam, Golam Rabiul; Rhaman, Khalilur
Research on planetary rovers has been active for a decade. Planetary rovers are usually designed for surf landing sites on other planets. Rovers with multiple functionalities make conducting interplanetary explorations in outer space practically possible. Most state-of-the-art interplanetary rovers are 6-wheelers and have difficulty performing a 360-degree rotation. In this research, the design and implementation process of a compact 4-wheeler rover named Mongol-Tori with robust capabilities has been approached which can be applied for the exploration of an unknown planet. Bearing in mind that the weight and volume increase the expenditure in space research, a lightweight modified suspension system has been proposed. An effective electronics system increases the run time and helps the driver of the rover to estimate the mission timeline in teleoperation mode. The vigorous design allows the onboard manipulator to complete any assistance task smoothly.
Metadata Only
A study on the development of an IoT-based health monitoring system for paralyzed (Quadriplegic) patients
(Institute of Electrical and Electronics Engineers Inc., 2023-01-01) Salim Rafid, Sk Tahmed; Miazee, Redwan Ahmed; Byzid, Md Kutubuddin; Anika, Afsana Anjum; Rhaman, Md Khalilur; Malek Azad, AKM Abdul
In this study, we explore the design of a modular, portable, non-invasive device for remote monitoring of temperature, heart rate, oxygen saturation, blood pressure and electrocardiogram (ECG) for quadriplegic patients with near-complete body paralysis. The developed system not only enables remote monitoring, through the interaction with the Internet of Things (loT) technology, of the patients but also reduces the caretakers' workload of continuous monitoring. The system is designed to be a low-cost and user-friendly solution that can be used at home/care homes for continuous monitoring. It also provides a cloud-based platform for monitoring and analyzing data, enabling remote access and providing real-time feedback to caretakers. The results of the system are validated using a few volunteers to demonstrate its accuracy and reliability. The cost-effective design, which includes the monitoring of five vitals, makes it desirable for Bangladesh's healthcare sector.
Metadata Only
Development of a brain-computer interface (BCI) for person with disabilities to control their wheelchair using brain waves
(Institute of Electrical and Electronics Engineers Inc., 2023-01-01) Rafid, S.K. Tahmed Salim; Miazee, Redwan Ahmed; Byzid, Md Kutubuddin; Anika, Afsana Anjum; Hossain, Syed Shakawat; Malek Azad, A.K.M. Abdul
Brain Computer Interface (BCI) is a system that identifies the functional intent, to interact with the surrounding environment, from the neuro-electric activity in the brain. This paper uses the BrainSense device, an EEG acquisitor, to acquire brain activity data from eye blink patterns. These are then processed and classified into different labels used as movement instructions for motors. The intention of this study is to develop a BCI system that allows quadriplegic individuals to operate a wheelchair without the need for physical/manual labor. Performance metrics include the accuracy between intended and translated execution. From the conducted study the developed wheelchair's performance significantly improved to 97.22% with 30 minutes of training prior to usage. The proposed system can have a long lasting impact on the healthcare industry in Bangladesh by aiding and enhancing the lives of people with disabilities.
Metadata Only
Automatic Bangla article content categorization using a hybrid deep learning model
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Mehedi, Md Humaion Kabir; Faruk, Kazi Omar; Rahman, Anika; Nessa, Iffatun; Zabin, Benozir; Nahar, Khairun; Iqbal, Shadab; Hossain, Md Sabbir; Rasel, Annajiat Alim
The process of categorizing or marking documents based on their content is known as document classification or categorization. Bangla is the world's fifth most widely spoken native language and the sixth most widely spoken language overall. Proper methods for automatic document classification are very limited. In this paper, we have introduced a hybrid deep learning model which is the combination of CNN-BiLSTM for automatic article categorization for Bangla language. We have also compared our model with different classifiers to measure its efficiency. Our proposed model acheived 83.48% accuracy.
Metadata Only
Recruitment scam detection using gated recurrent unit
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Nessa, Iffatun; Zabin, Benozir; Faruk, Kazi Omar; Rahman, Anika; Nahar, Khairun; Iqbal, Shadab; Hossain, Md Sabbir; Mehedi, Md Humaion Kabir; Rasel, Annajiat Alim
Recruitment advertisements in-cluding fake or fraudulent job advertisements is delivered to any job seekers and lead them to lose their money or personal information by frauds. With mass digitization and inter-net access scammers are coming up with new ideas and recruitment fraud is one of them. By sending false recruitment emails, SMS texts, fake websites, fake social media profiles, fake job posts, online recruiting services such as LinkedIn, or unsolicited emails purporting to be from well-known organizations are common methods of recruitment fraud. To identify fake or fraudulent job advertisements, we have pro-posed a one layer gated recurrent unit (GRU) model that can classify scams and real re-cruitments. We also evaluate the efficiency and achieved 93.51% of AUC score on Employment Scam Aegean Dataset (EMSCAD) dataset.
Metadata Only
Understanding sarcasm from reddit texts using supervised algorithms
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Hasnat, Fahim; Hasan, Md. Mazidul; Nasib, Abdullah Umar; Adnan, Ashik; Khanom, Nazifa; Islam, S M Mahsanul; Mehedi, Md Humaion Kabir; Iqbal, Shadab; Rasel, Annajiat Alim
The use of satirical or ironic language for conveying a message is referred to as Sarcasm. Social networks such as Reddit, Twitter, etc. usually uses Sarcasm. Reddit which is an American website contains social aggregations of news, ratings of the content and discussions. These resources, which include links, text articles, photographs, and videos, are published to the platform by registered users and can be voted up or down. Posts that cover topics related to books, cooking, pets, news, politics, movies, religions, science, sports, fitness, video games, music, and image-sharing are organized as 'communities' or 'subreddits'. The submissions receiving enough outvotes appear on the front page of the site and towards the top of the subreddit. This paper is about classifying a Reddit comment as sarcastic or non-sarcastic with the help of machine learning techniques. The dataset used in this study named as 'Sarcasm on Reddit' for classi-fication according to genre-based and have followed some basic steps using supervised machine learning and deep learning algorithms for the classification of texts including pre-processing, feature extraction and modeling. In this approach, we have achieved 71%, 76% and 70% accuracy for LSTM, CNN, and Logistic Regression algorithms respectively.
Metadata Only
Decentralized neural network based collaborative filtering for privacy concern recommendation systems
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Faruk, Kazi Omar; Rahman, Anika; Shusmita, Sanjida Ali; Awlad, Md Sifat Ibn; Das, Prasenjit; Mehedi, Md Humaion Kabir; Iqbal, Shadab; Rasel, Annajiat Alim
The growing concern about the privacy of user data is inspiring the development of new privacy preserving machine learning approaches. Decentralized federated learning is such a method which can handle privacy concerns effectively. It consists of servers and clients. Here, a machine learning model is distributed to a number of clients and the clients use their locally stored data to train the model and send the model to the server for aggregation. Here, the client only shares the model parameters with the server. The server receives thousands of locally trained models from clients and performs aggregation to define a global model. Only the model parameters such as weights and biases are being shared between the server and the client. In this paper, we have proposed a decentralized privacy preserving technique for neural collaborative filtering which is widely used in rating prediction for recommendation systems. Here each client receives an initial neural network based collaborative filtering model from the server and trains the model locally with its own data and only sends the model and parameters back to the server for aggregation. This method will eliminate privacy concerns in modern recommendation systems.
Metadata Only
COVID-19 classification from X-Ray images using 2D CNN
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Khan, Samiha; Islam, S M Mahsanul; Nasib, Abdullah Umar; Hasnat, Fahim; Hasan, Md. Mazidul; Mim, Sumaiya; Bin Sayed, Jawad; Mehedi, Md Humaion Kabir; Iqbal, Shadab; Rasel, Annajiat Alim
The coronavirus (COVID-19) detection has been a crucial task for researchers, scientists, health experts all across the world and everyone is trying together to find a solution to it. The X-rays images of lungs have become one of the most prevalent and effective procedures used by researchers to monitor COVID-19. Unfortunately, inspecting each case involves multiple radiology experts and time, which is one of the critical tasks in such an outbreak. In this paper, a deep learning approach, 2D convolutional neural networks (CNN) has been used to classify healthy and COVID-19 chest X-ray images. 'Curated Dataset for COVID-19 Posterior-Anterior Chest Radiography Images (X-Rays)' dataset has been used in this study. The major indicator of this study is the accuracy of the proposed model. The classification model, 2D CNN has achieved accuracy and f1-score of 0.96 and 0.95 respectively.
Metadata Only
Transliterated Bengali comment classification from social media
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Al Taawab, Abdullah; Tasnia, Lubaba; Dhar, Mondira; Mehedi, Md Humaion Kabir
In the era of technological advancement, the internet acts as an essential part of our daily life. People express their opinions on social media through different types of comments. In this paper, machine learning (ML) and deep learning (DL) models have been used to classify transliterated Bengali comments. Due to the lack of a large publicly available transliterated Bengali corpus, we have created our own dataset, consisting of 1,300 transliterated Bengali comments, which is publicly available in Mendeley Data. Moreover, we have applied several ML and DL algorithms, e.g., multinomial naive bayes (MNB), logistic regression (LR), linear SVM, decision Tree (DT), AdaBoost, random forest (RF), RBF SVM, gradient boosting, recurrent neural network (RNN), gated recurrent units (GRU), and long short-term memory (LSTM) for classifying comments. We have implemented different feature extraction techniques to compare the results. Among all these algorithms, logistic regression with countVectorizer performed best with 85.76% accuracy and 85.70% F1 score.
Metadata Only
A horizontal federated random forest for heart disease detection from decentralized local data
(Institute of Electrical and Electronics Engineers Inc., 2022-01-01) Jalal, Shafin Mahmud; Hasan, Md. Rezuwan; Haque, Md. Ashfaqul; Alam, Md. Golam Rabiul
In the modern world, reliable data is a thriving need in every sector. As the data increases, maintaining data privacy is also becoming a big concern. The healthcare sector is no different than that. Privacy in the healthcare sector is a topmost concern when sharing with other institutes. As data from a single healthcare institute is not always enough to get properly predicted outputs in machine learning approaches. There comes the idea of sharing data among multiple hospitals for having a more specified model with keeping the data details private. So, we have designed a model combining a federated central model and clients for the application of Federated Learning on heart disease patients' data. Here, we have implemented an approach for sharing only the model parameters among the clients and central in horizontal federated learning infused with random forest. At the evaluation of our model, we have come up with improved accuracy of 7.1, 2, and 6 percent respectively for the federated central and both clients.
Metadata Only
Performance analysis of machine learning approaches in diabetes prediction
(Institute of Electrical and Electronics Engineers Inc., 2021-01-01) Sakib S.; Yasmin N.; Tasawar, Ihtyaz Kader; Aziz A.; Bakr Siddique M.A.; Rahman Khan M.M.
Diabetes is a major chronic syndrome caused by a series of metabolic abnormalities in which blood glucose levels are abnormally high for an indeterminate amount of time. It influences various organs in the human body, resulting in a variety of complex diseases such as stroke, renal disease, pulmonary embolism, eyesight, and so on. Diabetes Disorders (DD) are presently one of the healthcare top causes of mortality. Predictive analytics in the health care system is a huge obstacle, but if accurate early prediction is achieved, the potential risk and degree of diabetes may be significantly decreased. Machine learning (ML) techniques are now used to analyze medical datasets at an earlier stage of life in keeping people safe. In this research, we utilized several ML approaches notably Logistic Regression, Decision Tree (DT), XGBoost, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Random Forest (RF) on PIMA Indian Diabetes Dataset in order to monitor and evaluate their performances in diabetes prediction. The performance of the various ML algorithms employed in this research suggests which algorithm is most suitable in diabetes prediction. It is observed that among all the models XGBoost had outperformed the other ML techniques with an accuracy of 80.73% while SVM was the second-best performing model with a classification accuracy of 80.21%. Thus, employing ML techniques, this study aims to assist doctors as well as clinicians in the early detection of diabetes.

Browse

Recent Submissions