Institution
Chittagong University of Engineering & Technology
Education•Chittagong, Bangladesh•
About: Chittagong University of Engineering & Technology is a education organization based out in Chittagong, Bangladesh. It is known for research contribution in the topics: Computer science & Renewable energy. The organization has 1200 authors who have published 1444 publications receiving 10418 citations. The organization is also known as: Engineering College, Chittagong & Bangladesh Institute of Technology, Chittagong.
Topics: Computer science, Renewable energy, Dielectric, Population, Solar cell
Papers published on a yearly basis
Papers
More filters
••
TL;DR: In this article, the salinity distribution in the estuarine channel network of a partially mixed estuary was investigated by quantifying salt intrusion length and stratification with response to tidal circulations.
Abstract: The salinity distribution in the estuarine channel network of a partially mixed estuary was investigated by quantifying salt intrusion length and stratification with response to tidal circ...
4 citations
••
01 Jan 2020
TL;DR: This work presents a method based on image processing technique for detecting driver drowsiness by considering eye blink rate and pressure sensor value from hand gloves and shows that it is performing as expected to detect driver drowsy.
Abstract: The drowsiness detection system is a non-evasive system which uses vision-based concepts. In this work, we present a method based on image processing technique for detecting driver drowsiness by considering eye blink rate and pressure sensor value from hand gloves. We applied the Haar Cascade Classifier and Viola–Jones algorithm to detect face and eyes. A camera is focused on the face of the subject from an adequate distance. The system first detects the driver’s face. Then it confirms the region of interest: eye. It then starts counting the blinking of the eye and hand grip pressure from the gloves. If the blinking rate is higher than the normal rate for a given period of time and the hand pressure value is lower than the threshold, the system confirms that the driver’s condition is drowsy. It then gives a warning signal to the driver. If the system cannot measure any blink for a fixed period of time and hand pressure value is extremely lower than the threshold, the proposed system confirms that the driver is sleepy. It then either slows down to the point of stopping the vehicle or gives a strong alarm to wake up the driver. Experimental evaluation is done with 5 mock-drivers of different age groups in a total of more than 4 h of recording time. Result shows that the system is performing as expected to detect driver drowsiness. Eye blinking rate together with hand grip pressure gives a better performing system making it an excellent candidate for future exploration in the field of automotive.
4 citations
•
01 Jan 2022
TL;DR: In this article, the authors proposed a method to determine an effective rate of the minority class over-sampling by which to maximize the performance of the machine learning model, which has achieved a top f1-score when the minority classes was over sampled by 30-45% of the majority class samples.
Abstract: Over-sampling is a resampling technique that has been designed to balance the imbalanced class distribution by duplicating samples of the minority class for a classification dataset. It is challenging to determine what rate of sample duplicating will be effective to maximize the model accuracy. In this research, we have proposed a method to determine an effective rate of the minority class over-sampling by which to maximize the performance of the machine learning model. We have used five over-sampling methods named Random over-sampling, SMOTE, SVMSMOTE, SMOTE Nominal, and Borderline SMOTE to evaluate the proposed method with five publicly available datasets. During the training period, we have over-sampled the minority class based on the majority class samples between the percentage ranges from 0 to 50%. Random Forest (RF) has been used as a machine learning classifier because its default hyperparameters already return great results. F1-score has been used as evaluation matrices because it is effective for imbalanced datasets. It has been seen that the proposed model has achieved a top f1-score when the minority class was over-sampled by 30–45% of the majority class samples.
4 citations
••
14 Dec 2020
TL;DR: In this article, an Isolation Forest Learning-based Outlier Detection Model for effective classifying cyber anomalies is presented. But, it is difficult to accurately model cyber threats since modern security databases contain large number of security features that could include Outliers.
Abstract: Cybersecurity has recently gained considerable interest in today’s security issues because of the popularity of the Internet-of-Things (IoT), the considerable growth of mobile networks, and many related apps. Therefore, detecting numerous cyber-attacks in a network and creating an effective intrusion detection system plays a vital role in today’s security. However, it is difficult to accurately model cyber threats since modern security databases contain large number of security features that could include Outliers. In this paper, we present an Isolation Forest Learning-Based Outlier Detection Model for effectively classifying cyber anomalies. In order to evaluate the efficacy of the resulting Outlier Detection model, we also use several conventional machine learning approaches, such as Logistic Regression (LR), Support Vector Machine (SVM), AdaBoost Classifier (ABC), Naive Bayes (NB), and K-Nearest Neighbor (KNN). The effectiveness of our propsoed Outlier Detection model is evaluated by conducting experiments on Network Intrusion Dataset with evaluation metrics such as precision, recall, F1-score, and accuracy. Experimental results show that the classification accuracy of cyber anomalies has been improved after removing outliers.
4 citations
••
01 Feb 2017TL;DR: Both text dependent and text independent procedure for speaker recognition is proposed using the responses of the model of the auditory system, which is applicable to real-time, text-dependent and independent speaker identification systems.
Abstract: The objective of this research is identifying a speaker from its voice regardless of the content. Speaker identification is a process to identify a speaker by voice biometrics. It is also known as voice recognition. In this study, both text dependent and text independent procedure for speaker recognition is proposed using the responses of the model of the auditory system. The robustness of the proposed model is also tested here by training the speech signals with the aid of Gaussian Mixture Model and testing by Probability Density Function. Performance tests conducted using the University Malaya and GRID database corpora have shown that this procedure has faster identification time and greater accuracy compared with traditional approaches, and so it is applicable to real-time, text-dependent and independent speaker identification systems.
4 citations
Authors
Showing all 1219 results
Name | H-index | Papers | Citations |
---|---|---|---|
Mohammad Rashed Iqbal Faruque | 29 | 288 | 2969 |
Ashraf Uz Zaman | 28 | 147 | 2704 |
Nusrat Jahan | 26 | 173 | 2127 |
M.M.K. Bhuiya | 22 | 45 | 1925 |
Iqbal H. Sarker | 20 | 100 | 1100 |
M. A. Ali | 20 | 87 | 1155 |
Scott Arthur | 19 | 106 | 1963 |
Mohammed Nazrul Islam | 17 | 124 | 905 |
M. G. Hafez | 16 | 60 | 735 |
Mohammad Mahbubur Rahman | 16 | 44 | 779 |
Khizir Mahmud | 15 | 54 | 851 |
Mohammad Alamgir Hossain | 15 | 62 | 638 |
J. U. Ahamed | 15 | 47 | 1151 |
Md. Mukter Hossain | 14 | 46 | 593 |
Kaushik Deb | 14 | 97 | 833 |