I am currently working as a faculty in Military Institute of Science and Technology in the department of Electrical, Electronic and Communication Engineering. Additionally, I am a part time research engineer at Visual Information Processing research lab at MIST.
I have completed my B.Sc. in Electrical, Electronic and Communication Engineering (EECE) from the Military Institute of Science and Technology (MIST), affiliated with Bangladesh University of Professionals (BUP), Dhaka, Bangladesh on January 2021. Download CV
2021   |   Joined in Visual Information Processing Lab (VIP) Research Lab |
2021   |   Joined as Faculty in MIST |
2021   |   Graduated from MIST and secured a position of 2nd in the entire department |
2020   |   Recipient of Dean’s List Scholarship for Outstanding Academic Results |
2019   |   1st Runner-up of IEEE Conference Project Presentation |
2019   |   Recipient of Dean’s List Scholarship for Outstanding Academic Results |
2018   |   Recipient of Dean’s List Scholarship for Outstanding Academic Results |
2017   |   Recipient of Dean’s List Scholarship for Outstanding Academic Results |
House No: 25-30
Plot # 30, Road # 7,
Line:N-1, N-2, Block-J
Eastern Housing, 2nd Phase, Mirpur, Dhaka North, Dhaka-1216
+8801537012420
Adjunct Faculty Military Institute of Science and Technology Course Taken
|
Research Engineer Visual Information Processing(VIP) Lab, MIST Projects
|
Industrial TraineeGrameenphone Ltd.Skills
|
Industrial TraineeBDcom online limitedSkills
|
Senior Mentor : App DevelopmentMIST Innovation ClubResponsibilities
|
Duration: |   4 Years (Jan 2017 - Jan 2021) |
Degree: |   Bachelor of Science (B.Sc.) |
Major: |   Communication |
CGPA: |   3.98/4.00 |
Rank: |   2nd in the entire University |
Recognition: |   Recipient of Dean's List of Honour, MIST Medal |
Department: |   Electrical, Electronic and Communication Engineering (EECE) |
Faculty: |   Electrical and Computer Engineering (ECE) |
Institute: |   Military Institute of Science and Technology (MIST) |
University: |   Bangladesh University of Professionals (BUP) |
My research interests and experiences are mostly in the machine learning and deep learning domain along with computer vision applications. In addition, my other research interests include : Explainable A.I., Human-Computer Interaction, Optimization Theory and Probabilistic Models.
Manual interpretation and classification of ECG signals lack both accuracy and reliability. These continuous time- series signals are more effective when represented as an image for CNN-based classification. A continuous Wavelet transform filter is used here to get corresponding images. In achieving the best result generic CNN architectures lack sufficient accuracy and also have a higher run–time. To address this issue, we propose an ensemble method of transfer learning-based models to classify ECG signals. In our work, two modified VGG-16 models and one InceptionResNetV2 model with added feature extracting layers and ImageNet weights are working as the backbone. After ensemble, we report an increase of 6.36% accuracy than previous MLP based algorithms. After 5-fold cross-validation with the Physionet dataset, our model reaches an accuracy of 99.98%.
Precise eye gaze detection has a multitude of real- life use cases such as the input mechanism for physically disabled persons, driver’s attention detection in vehicles, cheating detection in online exam, augmented reality, medical research and so on. Most of the applications need to support real-time functionality, thus the need for a fast and reliable method for eye gaze detection can be justified. In this research work, we propose a non-wearable and webcam-based eye-gaze detection method that offers multiple benefits in terms of accuracy, robustness, and reliability over existing solutions. We leveraged the latest innovation and breakthroughs in deep learning to construct a novel eye-gaze detection method that works using the live video feed from any modern webcam with acceptable frame rates for proper real-time applications. We achieved 99% validation accuracy in gaze prediction and 20 FPS on average in real-time applications such as mouse pointer control and scrolling.
With the advent of smart grids, accurate electric load forecasting has become more essential since it may assist power companies in improving load scheduling and reducing surplus energy output. Short term load forecasting (STLF) is gaining popularity owing to its utility in energy usage, demand- side management, energy storage, peak load forecasting and minimize electricity production costs. This study offers four artificial intelligence-based models to enhance 168-hours prediction accuracy. These models are long short term memory (LSTM), bidirectional LSTM (Bi-LSTM), Conv2D LSTM and Fbprophet. The models are trained with hourly energy consumption data of four years. After training and testing, it is depicted that bidirectional LSTM can predict more precisely than other models with an MAPE of 3.59. The MAPE of Conv2D LSTM, LSTM and Fbprophet are found 3.95, 4.91 and 7.75 accordingly. Since bidirectional LSTM utilizes the LSTM regular model twice, they usually have more accuracy than conventional LSTM. The use of bidirectional LSTM may thus make the demand response system more efficient.
As mmWave has a wide range of applications, it has drawn a significant amount of attention in recent years. It has already been introduced in the next generation wireless communication system. In practice, it shows some shortcomings and most of these are eliminated by introducing beamforming which utilizes the spatial diversity enabled by Massive MIMO. Still, there are a few challenges in designing an efficient system for highly mobile users and making sure proper coverage and reliability. In this research, a machine learning based coordinated beamforming technique has been explored that supports highly mobile applications in mmWave systems with massive antenna arrays. The optimization of the deep learning model itself can increase the system performance as well as reduce the computational time complexity. The purpose of this work was to optimize the deep learning model and recommend proper initialization method to maximize the system performance. We found that for Xavier normal intialization algorithm the effective achievable rate is highest for least amount of data.