BRAC University Institutional Repository
Preserving Knowledge, Advancing Research, Sharing Scholarship
A digital platform for collecting, preserving, and sharing BRAC University’s scholarly and institutional outputs.
Contribute to the Repository
Faculty members, students are invited to self-submit their research, theses, and scholarly output.
Recent Submissions
Machine fault diagnosis using EMD-gammatone texture representation and a lightweight self-attention SqueezeNet
(Institute of Electrical and Electronics Engineers Inc., 2024-01-01) Zabin M.; Binte Kabir, Anika Nahian; Kabir, Muhammad Khubayeeb; Choi H.J.; Uddin J.; Department of Computer Science and Engineering
Machine fault diagnosis involves the intricate process of detecting and isolating faults, which is particularly challenging due to various sources of noise and the complex nature of the faults. In automatic fault diagnosis, feature engineering becomes daunting because of the time-varying characteristics of the fault signals. As artificial intelligence continues to advance, deep learning models have been increasingly employed for machine fault classification. However, a significant limitation of state-of-the-art models is their high computational complexity, making them unsuitable for deployment on portable devices. In this paper, a lightweight fault diagnosis model is proposed that consists of a self-attention SqueezeN et architecture along with a hybrid texture representation technique using empirical mode decomposition (EMD) and gammatone-spectrogram (GS) filter. In the model, initially, the dominant signal is extracted from the ID audio fault signals by discarding lower intrinsic mode functions (IMFs) from EMD and then the dominant signals are converted to 2D texture maps applying the GS filter. Then, the generated texture maps are fed to the modified self-attention SqueezeNet classifier, with reduced model width and depth as input for training and validation. In the experimental evaluation, two public benchmark datasets- ToyADMOS and MIMII are used to validate the model. Finally, the model classifies the machine audio faults using the trained features. The experimental results demonstrated that the hybrid EMD-Gammatone texture imaging outperforms the other state-of-the-art methods like MFCC, Gammatone, and Hilbert Huang Transform with a self-attention (SA) based SqueezeNet architecture. The EMD-Gammatone spectrum-based feature extraction accurately detected the faults by exhibiting an accuracy of 89.32% and 96.46% for MIMII and ToyADMOS datasets, respectively. In addition, with the EMD-Gammatone Spectrogram images, the proposed model outperforms the conventional SqueezeNet and other state-of-the-art deep architectures with comparatively higher Precision, Recall, and FI scores. Furthermore, the proposed model gains reduced computational complexity due to 93.4% fewer trainable parameters of SqueezeN et than the conventional SqueezeN et model and the attention mechanism that focuses on important regions or features of the input.
A comparative study on Bengali speech sentiment analysis based on audio data
(Institute of Electrical and Electronics Engineers Inc., 2023-01-01) Shruti, Abanti Chakraborty; Rifat, Rakib Hossain; Kamal, Marufa; Alam, Md. Golam Rabiul; Department of Computer Science and Engineering
Sentiment analysis is one of the most researched areas for every language. Due to the rise of AI, the use of speech in every sector is rapidly growing so is the importance of Speech Sentiment Analysis. Despite being the seventh most spoken language in the world, Bengali speech sentiment analysis studies are not much enriched. This study compared the Bengali speech sentiment analysis using machine learning and CNN, LSTM, and Bi-LSTM models. We have used the SUBESCO and BanglaSER datasets for training our models where the KNN model outperformed other models with an accuracy of 90%. Later, we evaluated the performance of the models with our custom-made test dataset. Experimental results show that AdaBoost and Bi-LSTM model performed best with 45% accuracy. Moreover, to understand the feature effect on the output, we used the interpretable SHAP model in the ML model outcomes as they provide the best results allowing us to have an explainable advantage to determine the results.
Influence of cyber-victimization and other factors on depression and anxiety among university students in Bangladesh
(BioMed Central Ltd, 2023-12-01) Rahman, Tareq; Hossain, Md. Mahin; Bristy, Nurun Nahar; Hoque, Md. Zahidul; Hossain, Md. Moyazzem; BRAC James P Grant School of Public Health
Background and objectives: Cyber-victimization is closely linked with mental health problems such as anxiety, depression, etc., and has become a growing concern among university students in Bangladesh. In the era of globalization, smart gadgets, the internet, and other online resources are readily available, and these tools and devices have now become the primary method for cyberbullying. The authors aim to explore the impacts of cyber-victimization and other factors on anxiety and depression among university students in Bangladesh. Methods: The primary data for this cross-sectional study were collected using a well-structured questionnaire. This study employs three widely used scales such as cyberbullying inventory, general anxiety disorder-7 (GAD-7), and patient health questionnaire-9 (PHQ-9). Descriptive statistics and multivariable logistic regression analyses are carried out to identify the factors associated with depression and anxiety among university students in Bangladesh. Results: Findings depict that the prevalence of depression and anxiety among university students was 52.5% and 44.0%, respectively. Depressed respondents were considerably more likely to have difficulty sleeping (p < 0.001), spend more time on social media (p = 0.002), have suicidal thoughts (p < 0.001), and have a high cyber-victimization score (p < 0.001) compared to non-depressed respondents. In comparison with non-anxious respondents, anxious respondents were significantly more likely to: have sleeping difficulties (p < 0.001); spend more time on social media (p = 0.031); have suicidal thinking (p < 0.001); and have a comparatively high cyber-victimization score (p < 0.001). Multivariable logistic regression analysis identified that a one-unit increase in the cyber-victimization score results in a 1.24 times higher chance of experiencing depression (AOR: 1.24, 95% CI 1.17–1.31, p < 0.001), and a one-unit increase in the cyber-victimization score results in a 1.23 times higher chance of experiencing anxiety (AOR: 1.23, 95% CI 1.17–1.30, p < 0.001). Conclusion: University students are struggling with cyberbullying, which can lead to depression and anxiety levels. Promoting more cyberbullying awareness is necessary since failing to do so could result in a sustained or increased prevalence of anxiety and depression levels among students, which could have disastrous repercussions. © 2023, The Author(s).
Microfluidic simulation framework for point-of-care diagnostics: modeling advection-diffusion-reaction transport in lab-on-chip systems
(Institute of Electrical and Electronics Engineers Inc., 2025-01-01) Shitab T.A.; Ibnat Oni, Anika; Hossain Emon M.E.; Rayan Rahat R.; Mim A.A.; Fariya Auishe P.; Department of Biotechnology
Point-of-care diagnostics (POCD) rely heavily on microfluidic lab-on-chip systems, yet their design process is often hindered by trial-and-error prototyping and inconsistent reproducibility. Efficient diagnostic performance requires a clear understanding of how advection, diffusion, and biochemical reactions jointly govern analyte transport at the microscale. This work presents a lightweight numerical framework for modeling solute behavior in microfluidic channels using the advection-diffusion-reaction equation. A finite-difference solver was applied to a 5 mm × 100 ?m channel under physiologically relevant flow conditions, comparing three regimes: diffusion-only, advection + diffusion, and advection + diffusion + first-order reaction. Results show that high Péclet numbers (Pe ? 1000) drive plug-like transport with minimal lateral mixing, while Damköhler numbers near unity (Da ? 1) produce significant axial decay due to reaction kinetics. Visualizations of concentration fields and breakthrough curves demonstrate how Pe and Da directly influence sensor placement, analyte preservation, and diagnostic sensitivity. The framework provides a transparent and computationally efficient tool for early-stage device design, offering practical guidance for optimizing flow control, channel geometry, and capture-surface behavior in lab-on-chip diagnostics.
Water, sanitation and hygiene (WASH) practices and deworming improve nutritional status and anemia of unmarried adolescent girls in rural Bangladesh
(BioMed Central Ltd, 2023-12-01) Jolly, Saira Parveen; Roy Chowdhury, Tridib; Sarker, Tanbi Tanaya; Afsana, Kaosar; BRAC James P Grant School of Public Health
Background: In Bangladesh, undernutrition and anemia are more occurrent among adolescent girls. BRAC, the largest non-governmental organization (NGO), has been implementing a community-based nutrition education service package targeting adolescent girls for reducing their undernutrition and anemia. Objective: We aimed to explore the underlying factors associated with nutritional status and anemia among adolescent girls under the BRAC nutrition program areas to improve their existing intervention package. Methodology: We conducted a cross-sectional and comparative study in 2016, in 24 upazilas of Bogra, Barguna, Comilla, Dinajpur, Feni, Jessore, and Meherpur districts where the BRAC nutrition program was implemented while the remaining 27 upazilas of those districts were selected as comparison area. We followed a multistage cluster random sampling for selecting 1620 unmarried adolescent girls aged 10–19 years for interviewing in the intervention and comparison areas. Data were collected on socio-demographic information, dietary intake, morbidity, water, sanitation, and hygiene (WASH) practice, anthropometry, and serum hemoglobin (Hb) level by using a pre-structured questionnaire. The nutritional status of the adolescent girls was expressed as height-for-age Z (HAZ) and body mass index-for-age Z (BMIZ) score, while anemia referred to the serum Hb at the level of below 12 g/dl for adolescent girls. All statistical analyses were done in STATA version 17 (Chicago Inc.). Findings: The prevalence of stunting (22.9% vs. 22.5%), thinness (12% vs. 14%), and anemia (34.5% vs. 37.3%) exhibited similarities between the intervention and comparison regions. Stunting and thinness were predictors for each other for this population group. Our findings indicated that adolescent girls who were not washing hands with soap after defecation were likely to be stunted [AOR 1.51 (95% CI 1.12–2.04)], and who did not utilize sanitary latrines had an increased likelihood of being thin [AOR 2.38 (95% CI 1.11–5.08)]. Conversely, those who did not watch television [AOR 1.69 (95% CI 1.12–2.56)] and did not have deworming tablets [AOR 1.33 (95% CI 1.07–1.64)] in the 6 months leading up to the interview had a 69% and 33% higher probability of being anemic, respectively. Conclusion: For sustainable improvement in the undernutrition and anemia of adolescent girls, integration of WASH, consistent administration of deworming tablets and broadcasting awareness programs through television are urgent to scale up the nutrition intervention programs in similar settings like Bangladesh. © 2023, The Author(s).