Mohammad Salah Uddin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Mohammad Salah Uddin has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Aoyan Barua
Dept. of Computer Science and Engineering, East West University, Dhaka, Bangladesh
2 shared publicationsJannat Binta Alam
Department of Computer Science and Engineering, East West University, Bangladesh
2 shared publicationsShamim Ripon
2 shared publicationsAlcor Lab
Department of Computer Control and Management Engineering “A. Rubert”, Sapienza University of Rome, Rome, Italy
1 shared publication- Ali Akbar Aurnab
East West University
1 shared publication - Anika Tabassum Nafisa
East West University
1 shared publication
Latest Publications
- Open Access
Quantum-inspired multi-objective feature selection for Cross-Project software Defect Prediction
- 30 June 2026
- Published by IGI Global Scientific Publishing
- p. 45-90
Problem to Proposal
Capstone projects need to be mapped to real-world problems. In this chapter, we will discuss a structured way for students and faculty to choose meaningful problems and transform them into clear and effective capstone proposals. It begins by looking at societal challenges, industry needs, and community-based opportunities. It then shows some tools for narrowing down large, broad ideas into focused research or design-type questions. Students learn how to match their own interests with academic goals and ensure that their final work actually returns some real value to stakeholders. This chapter highlights why a clear purpose (the “why”) matters and gives strategies for actually running a project properly while avoiding issues like no scope or fuzzy objectives. Readers get walked through how to build an engaging capstone proposal: from writing a solid problem statement. Also know how to set SMART goals, mapping out deliverables, and doing early communication with advisors and end-users. There is also several sample proposal with annotations, so the practical ideas do not just stay theoretical but can be seen in action. To make sure projects are actually doable, this chapter includes a feasibility checklist. The checklist that touches on time management, team capacity, resources, and ethical stuff (like research involving human or animal subjects). There is also a dedicated “Mentor's Corner” section where faculty or supervisors share practical advice, helping students polish their rough ideas without mentors having to run the whole project for them. After going through this chapter, students will have a repeatable framework for turning curiosity or concern into a clear, actionable project proposal. This sets up a solid foundation for a meaningful and successful capstone experience.
- 30 June 2026
- Published by IGI Global Scientific Publishing
- p. 45-90
Problem to Proposal
Capstone projects need to be mapped to real-world problems. In this chapter, we will discuss a structured way for students and faculty to choose meaningful problems and transform them into clear and effective capstone proposals. It begins by looking at societal challenges, industry needs, and community-based opportunities. It then shows some tools for narrowing down large, broad ideas into focused research or design-type questions. Students learn how to match their own interests with academic goals and ensure that their final work actually returns some real value to stakeholders. This chapter highlights why a clear purpose (the “why”) matters and gives strategies for actually running a project properly while avoiding issues like no scope or fuzzy objectives. Readers get walked through how to build an engaging capstone proposal: from writing a solid problem statement. Also know how to set SMART goals, mapping out deliverables, and doing early communication with advisors and end-users. There is also several sample proposal with annotations, so the practical ideas do not just stay theoretical but can be seen in action. To make sure projects are actually doable, this chapter includes a feasibility checklist. The checklist that touches on time management, team capacity, resources, and ethical stuff (like research involving human or animal subjects). There is also a dedicated “Mentor's Corner” section where faculty or supervisors share practical advice, helping students polish their rough ideas without mentors having to run the whole project for them. After going through this chapter, students will have a repeatable framework for turning curiosity or concern into a clear, actionable project proposal. This sets up a solid foundation for a meaningful and successful capstone experience.
- 12 June 2026
- Published by IGI Global Scientific Publishing
- p. 57-90
Generation Z and Boundless Work
The growth of freelancing and online platform-based work has significantly changed how Generation Z in Bangladesh see employment and job security. For many young people, work is no longer limited to offices or local job markets. Now, digital platforms give them the opportunity to work for clients worldwide. They can do this while staying at home, in a café, or even in shared rooms. This chapter looks at how Gen Z participate in this borderless digital work. It also explores how it is shaping work culture in a Global South setting. Based on interviews with young freelancers, content creators, and remote workers, the chapter discusses why digital work is becoming more and more popular among this generation. Many participants say they choose freelancing because formal jobs are limited and the recruitment process is very strict. They also mention that they appreciate freedom, flexible timing, and less control from bosses. Earning foreign currency and being judged by their skills instead of their degree also makes platform work attractive. For Gen Z, digital work is not just about making money. It is also a way to show their abilities and independence. However, the chapter also points out the unstable side of platform work. Income is not fixed, and there is no real job security or social protection. Changes in algorithms and unpaid work are common problems faced by these young workers. To deal with these issues, they use informal methods like learning multiple skills, building peer groups, and promoting themselves online. Access to devices, internet, and safe workspace is not equal for everyone. Gender and location still play a big role in who can benefit more from this digital economy. By including Bangladeshi Gen Z in the wider discussion on digital economy and youth labour, this chapter argues that freelancing is more than just a short-term solution. It shows a shift in how young people think about work and success. Bangladeshi Gen Z are not just following global systems. They are also adjusting and negotiating these systems in their own way, in today's digital age.
- 12 June 2026
- Published by IGI Global Scientific Publishing
- p. 57-90
Generation Z and Boundless Work
The growth of freelancing and online platform-based work has significantly changed how Generation Z in Bangladesh see employment and job security. For many young people, work is no longer limited to offices or local job markets. Now, digital platforms give them the opportunity to work for clients worldwide. They can do this while staying at home, in a café, or even in shared rooms. This chapter looks at how Gen Z participate in this borderless digital work. It also explores how it is shaping work culture in a Global South setting. Based on interviews with young freelancers, content creators, and remote workers, the chapter discusses why digital work is becoming more and more popular among this generation. Many participants say they choose freelancing because formal jobs are limited and the recruitment process is very strict. They also mention that they appreciate freedom, flexible timing, and less control from bosses. Earning foreign currency and being judged by their skills instead of their degree also makes platform work attractive. For Gen Z, digital work is not just about making money. It is also a way to show their abilities and independence. However, the chapter also points out the unstable side of platform work. Income is not fixed, and there is no real job security or social protection. Changes in algorithms and unpaid work are common problems faced by these young workers. To deal with these issues, they use informal methods like learning multiple skills, building peer groups, and promoting themselves online. Access to devices, internet, and safe workspace is not equal for everyone. Gender and location still play a big role in who can benefit more from this digital economy. By including Bangladeshi Gen Z in the wider discussion on digital economy and youth labour, this chapter argues that freelancing is more than just a short-term solution. It shows a shift in how young people think about work and success. Bangladeshi Gen Z are not just following global systems. They are also adjusting and negotiating these systems in their own way, in today's digital age.
- Open Access
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Engineering
- Vol. 2 (1) , 015013
Hybrid quantum-classical classification of waste using the TrashNet dataset
This study presents a hybrid quantum-classical model for classifying waste images from the TrashNet dataset. TrashNet is a publicly accessible dataset of waste classifications. A classical convolutional neural network (CNN) is used to extract and reduce image features. These features are then passed into a variational quantum circuit with 4 qubits and a circuit depth of 3. The quantum layer performs the final classification. The system is implemented using PyTorch and PennyLane and trained on a simulated quantum backend. The hybrid model achieves 87.4% test accuracy using fewer than 300 trainable parameters. In comparison, a baseline classical CNN with similar accuracy required more than 11 260 parameters. Furthermore, the quantum model provides 1.6 times faster estimation and converges more faster during training. These results indicate that quantum models can reduce parameter calculations and improve computational efficiency. This work highlights the potential of quantum machine learning for low-resource, sustainable AI applications.
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Engineering
- Vol. 2 (1) , 015013
Open AccessHybrid quantum-classical classification of waste using the TrashNet dataset
This study presents a hybrid quantum-classical model for classifying waste images from the TrashNet dataset. TrashNet is a publicly accessible dataset of waste classifications. A classical convolutional neural network (CNN) is used to extract and reduce image features. These features are then passed into a variational quantum circuit with 4 qubits and a circuit depth of 3. The quantum layer performs the final classification. The system is implemented using PyTorch and PennyLane and trained on a simulated quantum backend. The hybrid model achieves 87.4% test accuracy using fewer than 300 trainable parameters. In comparison, a baseline classical CNN with similar accuracy required more than 11 260 parameters. Furthermore, the quantum model provides 1.6 times faster estimation and converges more faster during training. These results indicate that quantum models can reduce parameter calculations and improve computational efficiency. This work highlights the potential of quantum machine learning for low-resource, sustainable AI applications.
- 7 May 2026
- Published by Springer Nature
- p. 241-258
Smart Living for Seniors: Bridging Technology and Elderly Well-Being
- 7 May 2026
- Published by Springer Nature
- p. 241-258
Smart Living for Seniors: Bridging Technology and Elderly Well-Being
- Open Access
- 5 January 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Adaptive motion planning for legged robots in unstructured terrain using deep reinforcement learning
Legged robots are ideal for navigating unstructured terrain. Unlike wheeled robot platforms, legged robots maintain mobility over rocks, slopes, and uneven surfaces. However, motion planning in such environments remains challenging. The robot must handle unstable ground, limited sensing, and complex body dynamics. Classical control methods often fail to adapt when terrain and friction change unexpectedly. This paper presents an adaptive locomotion framework for quadrupedal robots. The controller is trained entirely in simulation using deep reinforcement learning (DRL). The DRL is trained using proximal policy optimization (PPO) method. During training, the robot observes joint states, body attitude, and local terrain height information. Training follows a curriculum mechanism. The robot starts on flat terrain. The difficulty gradually increases with slopes, height noise, rough surfaces, and friction variation. The reward function encourages forward progress, stability, low foot slip, and energy-efficient motion. Training occurs in Webots with wide domain randomization ranges. Policies learn to select footholds, regulate body attitude, and minimize slip without explicit terrain labels or gait scripts. Several behaviors emerge through learning. These include wider stance on slopes, increased step height on rocks, and hip abduction for balance recovery. To ensure robustness, we apply a multi-step validation process. These include cross-simulator evaluation, domain randomization Monte Carlo tests, physics consistency tests and adversarial disturbances. The controller achieves 94.6% success in Webots and 91.2% in PyBullet without retraining. It recovers from external pushes within 1.6 seconds. Results confirm that the proposed method learns stable, energy-efficient walking with minimal foot slippage across friction coefficients \(\mu = 0.1 to 1.0\). This capability supports deployment in search-and-rescue, off-road mobility, and exploration tasks in unknown terrain.- 5 January 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessAdaptive motion planning for legged robots in unstructured terrain using deep reinforcement learning
Legged robots are ideal for navigating unstructured terrain. Unlike wheeled robot platforms, legged robots maintain mobility over rocks, slopes, and uneven surfaces. However, motion planning in such environments remains challenging. The robot must handle unstable ground, limited sensing, and complex body dynamics. Classical control methods often fail to adapt when terrain and friction change unexpectedly. This paper presents an adaptive locomotion framework for quadrupedal robots. The controller is trained entirely in simulation using deep reinforcement learning (DRL). The DRL is trained using proximal policy optimization (PPO) method. During training, the robot observes joint states, body attitude, and local terrain height information. Training follows a curriculum mechanism. The robot starts on flat terrain. The difficulty gradually increases with slopes, height noise, rough surfaces, and friction variation. The reward function encourages forward progress, stability, low foot slip, and energy-efficient motion. Training occurs in Webots with wide domain randomization ranges. Policies learn to select footholds, regulate body attitude, and minimize slip without explicit terrain labels or gait scripts. Several behaviors emerge through learning. These include wider stance on slopes, increased step height on rocks, and hip abduction for balance recovery. To ensure robustness, we apply a multi-step validation process. These include cross-simulator evaluation, domain randomization Monte Carlo tests, physics consistency tests and adversarial disturbances. The controller achieves 94.6% success in Webots and 91.2% in PyBullet without retraining. It recovers from external pushes within 1.6 seconds. Results confirm that the proposed method learns stable, energy-efficient walking with minimal foot slippage across friction coefficients \(\mu = 0.1 to 1.0\). This capability supports deployment in search-and-rescue, off-road mobility, and exploration tasks in unknown terrain. - Open Access
- 1 January 2026
- Published by Wiley in Applied Computational Intelligence and Soft Computing
- Vol. 2026 (1)
Software Defect Prediction With Quantum‐Inspired Feature Selection
Accurate software defect prediction is crucial for reducing maintenance costs and improving system reliability. Effective feature selection ensures that models focus on the most informative metrics while avoiding overfitting and excessive complexity. We introduce a quantum‐inspired evolutionary algorithm (QIEA) that encodes candidate feature subsets using probability amplitudes and then iteratively refines them through adaptive rotation operations inspired by quantum mechanics. Our approach simultaneously optimizes prediction quality and model simplicity by balancing F1‐score maximization with subset size reduction. We evaluate QIEA on four public benchmarks—NASA PC1, CM1, and PROMISE KC1, JM1—using both within‐project cross‐validation and leave‐one‐project‐out protocols. Across all scenarios, QIEA delivers a consistent absolute improvement of six percentage points in F1‐score over classical genetic algorithms, while reducing the number of selected features by 30–40 percent. Hypervolume analysis confirms that our method achieves superior tradeoffs between accuracy and compactness. The experimental setup and parameter settings are described in detail throughout this paper.
- 1 January 2026
- Published by Wiley in Applied Computational Intelligence and Soft Computing
- Vol. 2026 (1)
Open AccessSoftware Defect Prediction With Quantum‐Inspired Feature Selection
Accurate software defect prediction is crucial for reducing maintenance costs and improving system reliability. Effective feature selection ensures that models focus on the most informative metrics while avoiding overfitting and excessive complexity. We introduce a quantum‐inspired evolutionary algorithm (QIEA) that encodes candidate feature subsets using probability amplitudes and then iteratively refines them through adaptive rotation operations inspired by quantum mechanics. Our approach simultaneously optimizes prediction quality and model simplicity by balancing F1‐score maximization with subset size reduction. We evaluate QIEA on four public benchmarks—NASA PC1, CM1, and PROMISE KC1, JM1—using both within‐project cross‐validation and leave‐one‐project‐out protocols. Across all scenarios, QIEA delivers a consistent absolute improvement of six percentage points in F1‐score over classical genetic algorithms, while reducing the number of selected features by 30–40 percent. Hypervolume analysis confirms that our method achieves superior tradeoffs between accuracy and compactness. The experimental setup and parameter settings are described in detail throughout this paper.
- 5 December 2025
- Published by IGI Global Scientific Publishing
- p. 41-78
A Drone System With Vison for Fishing Activity Detection
Illegal, unreported, and unregulated (IUU) fishing is a major threat to marine ecosystems. The local economies also affected by IUU. Traditional methods of monitoring fishing activity are often costly, labor-intensive, and only cover limited areas. To address these limitations, this study introduces a compact, low-cost aerial surveillance system that detects fishing boats and fishing gear using drone-mounted vision system. The system consists of a Raspberry Pi 4 (8GB ram version), coupled with a high-resolution Pi Camera and integrated with a LinkIt ONE board for GPS positioning and GPRS-based communication provider. A lightweight object detection model (YOLOv8n) is deployed on the Raspberry Pi to identify boats and fishing gear from aerial video footage. Each detection is geo-tagged using real-time GPS data and transmitted to a remote server/cloud. The LinkIt ONE's GPRS (cellular) module provides data transmission connectivity. It also supplies GPS data to the system for accurate positioning. This system is designed for efficiency and runs on low power. The computational overhead of the system is very low. It supports near-real-time detection at about 3 frames per second during daylight. The solution is suitable for coastal areas and protected marine areas. It can be used in inland waters (rivers or large fisheries) if necessary. The system supports both enforcement agencies and fisheries conservation programs. The vision-based system can effectively detect fishing activity. Overall experimental results show the reliable performance of the system. The overall cost is much lower than traditional aerial surveillance methods.
- 5 December 2025
- Published by IGI Global Scientific Publishing
- p. 41-78
A Drone System With Vison for Fishing Activity Detection
Illegal, unreported, and unregulated (IUU) fishing is a major threat to marine ecosystems. The local economies also affected by IUU. Traditional methods of monitoring fishing activity are often costly, labor-intensive, and only cover limited areas. To address these limitations, this study introduces a compact, low-cost aerial surveillance system that detects fishing boats and fishing gear using drone-mounted vision system. The system consists of a Raspberry Pi 4 (8GB ram version), coupled with a high-resolution Pi Camera and integrated with a LinkIt ONE board for GPS positioning and GPRS-based communication provider. A lightweight object detection model (YOLOv8n) is deployed on the Raspberry Pi to identify boats and fishing gear from aerial video footage. Each detection is geo-tagged using real-time GPS data and transmitted to a remote server/cloud. The LinkIt ONE's GPRS (cellular) module provides data transmission connectivity. It also supplies GPS data to the system for accurate positioning. This system is designed for efficiency and runs on low power. The computational overhead of the system is very low. It supports near-real-time detection at about 3 frames per second during daylight. The solution is suitable for coastal areas and protected marine areas. It can be used in inland waters (rivers or large fisheries) if necessary. The system supports both enforcement agencies and fisheries conservation programs. The vision-based system can effectively detect fishing activity. Overall experimental results show the reliable performance of the system. The overall cost is much lower than traditional aerial surveillance methods.
- 24 November 2025
- Published by Wiley in Computer Applications in Engineering Education
- Vol. 33 (6)
Analyzing and Predicting Student Performance in Discrete Mathematics Using Supervised Learning Algorithms
Discrete Mathematics is an important and challenging course for computer science and engineering students. It includes topics, such as logic, sets, proofs, number theory, graphs, trees, computation, relations, functions, and basic algorithmic concepts. These topics require strong analytical reasoning and consistent effort. As a result, many students find this course challenging to perform well. The aim of this study is to predict student performance in a Discrete Mathematics course at a reputed private university located in Bangladesh. Data were collected from both course instructors and students during the spring and summer semester of 2025. Instructors provided academic records, such as attendance, quizzes, assignments, and midterm scores. Students provided additional information, which included daily study time, subject interests, and use of learning platforms. The final data set included records for 240 students. K ‐means clustering with the Davies–Bouldin method was used to group similar students. Then, four machine learning (ML) models were trained and tested: Support Vector Machine (SVM), Decision Tree, K ‐Nearest Neighbors, and Naïve Bayes. The models were implemented using Python's scikit‐learn library, with stratified sampling and fivefold cross‐validation. Among the models, SVM achieved the highest accuracy of 96% after parameter tuning. Naïve Bayes had the lowest accuracy due to the assumption of feature independence. Key predictors of performance included mean score, attendance, and daily study hours. Findings show that ML can help instructors identify at‐risk students early, provide focused academic support, and improve learning outcomes. While the results are promising, the study is limited by sample size and does not include psychological or emotional factors. Future work will explore larger data sets and apply interpretable Artificial Intelligence techniques for better model transparency.
- 24 November 2025
- Published by Wiley in Computer Applications in Engineering Education
- Vol. 33 (6)
Analyzing and Predicting Student Performance in Discrete Mathematics Using Supervised Learning Algorithms
Discrete Mathematics is an important and challenging course for computer science and engineering students. It includes topics, such as logic, sets, proofs, number theory, graphs, trees, computation, relations, functions, and basic algorithmic concepts. These topics require strong analytical reasoning and consistent effort. As a result, many students find this course challenging to perform well. The aim of this study is to predict student performance in a Discrete Mathematics course at a reputed private university located in Bangladesh. Data were collected from both course instructors and students during the spring and summer semester of 2025. Instructors provided academic records, such as attendance, quizzes, assignments, and midterm scores. Students provided additional information, which included daily study time, subject interests, and use of learning platforms. The final data set included records for 240 students. K ‐means clustering with the Davies–Bouldin method was used to group similar students. Then, four machine learning (ML) models were trained and tested: Support Vector Machine (SVM), Decision Tree, K ‐Nearest Neighbors, and Naïve Bayes. The models were implemented using Python's scikit‐learn library, with stratified sampling and fivefold cross‐validation. Among the models, SVM achieved the highest accuracy of 96% after parameter tuning. Naïve Bayes had the lowest accuracy due to the assumption of feature independence. Key predictors of performance included mean score, attendance, and daily study hours. Findings show that ML can help instructors identify at‐risk students early, provide focused academic support, and improve learning outcomes. While the results are promising, the study is limited by sample size and does not include psychological or emotional factors. Future work will explore larger data sets and apply interpretable Artificial Intelligence techniques for better model transparency.
- 9 October 2025
- Published by IGI Global Scientific Publishing
- p. 341-400
IoT-Enabled Wearable Device for Stroke Rehabilitation at Home
This article presents a LoRa-based health monitoring device designed for rehabilitation patients. The system combines body surface EMG sensing with inertial motion tracking to record muscle activation, joint angle, and repetition count. A wearable unit transmits processed features over LoRa (Long Range network) to a cloud server. The cloud is a place where data is stored. The cloud provides the endpoint for visualization. Patients and clinicians access the data via web and mobile dashboards that display real-time EMG graphs, angle observations, session history, and feedback. Signal preprocessing includes bandpass filtering, rectification, RMS calculation, and normalization. Experiments showed EMG error < 6%, angle error < 3°, and rep detection F1 ≈ 0.96. LoRa communication achieved > 90% packet delivery indoors and outdoors with latency < 500 ms. Battery life exceeded 30 hours of continuous use. Case studies confirmed utility in hospitals, homes, clinics, and rural centers.
- 9 October 2025
- Published by IGI Global Scientific Publishing
- p. 341-400
IoT-Enabled Wearable Device for Stroke Rehabilitation at Home
This article presents a LoRa-based health monitoring device designed for rehabilitation patients. The system combines body surface EMG sensing with inertial motion tracking to record muscle activation, joint angle, and repetition count. A wearable unit transmits processed features over LoRa (Long Range network) to a cloud server. The cloud is a place where data is stored. The cloud provides the endpoint for visualization. Patients and clinicians access the data via web and mobile dashboards that display real-time EMG graphs, angle observations, session history, and feedback. Signal preprocessing includes bandpass filtering, rectification, RMS calculation, and normalization. Experiments showed EMG error < 6%, angle error < 3°, and rep detection F1 ≈ 0.96. LoRa communication achieved > 90% packet delivery indoors and outdoors with latency < 500 ms. Battery life exceeded 30 hours of continuous use. Case studies confirmed utility in hospitals, homes, clinics, and rural centers.