Rajesh Kumar

Dr. Rajesh Kumar

University of Milano-Bicocca, Milan, Lombardy, Italy
orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 April 2025 in Information
Verified
Journal Article
Open access
PDF file
Rajesh KumarRajesh Kumar
Rajesh Kumar
GCGIACOMO CORVISIERI
GIACOMO CORVISIERI
TFTullio Flavio Fici
Tullio Flavio Fici
Syed Ibrar HussainSyed Ibrar Hussain
Syed Ibrar Hussain
DTDomenico Tegolo
Domenico Tegolo
CVCesare Valenti
Cesare Valenti
Facial expressions reflect psychological states and are crucial for understanding human emotions. Traditional facial expression recognition methods face challenges in real-world healthcare applications due to variations in facial structure, lighting conditions and occlusion. We present a methodology based on transfer learning with the pre-trained models VGG-19 and ResNet-152, and we highlight dataset-specific preprocessing techniques that include resizing images to 124 × 124 pixels, augmenting the data and selectively freezing layers to enhance the robustness of the model. This study explores the application of deep learning-based facial expression recognition in healthcare, particularly for remote patient monitoring and telemedicine, where accurate facial expression recognition can enhance patient assessment and early diagnosis of psychological conditions such as depression and anxiety. The proposed method achieved an average accuracy of 0.98 on the CK+ dataset, demonstrating its effectiveness in controlled environments. However performance varied across datasets, with accuracy rates of 0.44 on FER2013 and 0.89 on JAFFE, reflecting the challenges posed by noisy and diverse data. Our findings emphasize the potential of deep learning-based facial expression recognition in healthcare applications while underscoring the importance of dataset-specific model optimization to improve generalization across different data distributions. This research contributes to the advancement of automated facial expression recognition in telemedicine, supporting enhanced doctor–patient communication and improving patient care.

1 January 2023 in International Journal of Advanced Computer Science and Applications
Verified
Journal Article
Open access
PDF file
KLKaiyu Li
Kaiyu Li
XGXiugang Gong
Xiugang Gong
MTMuhammad Tahir
Muhammad Tahir
TWTao Wang
Tao Wang
Rajesh KumarRajesh Kumar
Rajesh Kumar
In the Automated Guided Vehicle (AGV) warehouse automatic guided vehicle system, the path planning algorithm for intelligent logistics vehicles is a key factor to ensure the stable and efficient operation of the system. However, the existing planning algorithms have problems such as single designing a route and the inability to intelligently evade moving barriers. The academic community has proposed various solutions to these problems, although they have improved the efficiency and quality of path planning to some extent, they have not completely solved problems such as poor safety in planning, the high number of path inflection points, poor path smoothness, easily getting stuck in deadlocks, and have not fully considered the running cost and practical implementation difficulty of algorithms. To address these issues, the article deeply researched traditional A* scheme and Dynamic Window Approach (DWA) technology and proposed designing a route method according to the fusion of the A* algorithm and DWA technology. The algorithm improved the A algorithm by introducing a sub-node optimization algorithm to solve problems for instance poor global path planning safety and easy deadlock. Moreover, the algorithm reduced the amount of global route reversal locations and increased path consistency by improving the evaluation function and removing redundant points of the A algorithm. Finally, by integrating the DWA algorithm, the intelligent logistics vehicle achieved dynamic obstacle avoidance capabilities for moving objects in the real world. Our simulations-based results on MATLAB framework show that the algorithm significantly improves path smoothness, path length, path planning time, and environmental adaptability compared to traditional algorithms, and basically meets the path planning requirements of the AGV system for intelligent logistics vehicles.

22 November 2020 in Photonics
Verified
Journal Article
Open access
PDF file
KMKhalid Hussain Mohammadani
Khalid Hussain Mohammadani
RBRIZWAN ASLAM BUTT
RIZWAN ASLAM BUTT
Kamran Ali MemonKamran Ali Memon
Kamran Ali Memon
FHFayaz Hassan
Fayaz Hassan
Abdul MajeedAbdul Majeed
Abdul Majeed
Rajesh KumarRajesh Kumar
Rajesh Kumar
The combination of a high-speed wireless network with passive optical network technologies has led to the evolution of a modern integrated fiber wireless (FiWi) access network. Compared to broadband wireless networks, the FiWi network offers higher bandwidth with improved reliability and reduced maintenance costs due to the passive nature of passive optical network (PON). Since the quality of service (QoS) is a baseline to deploy high-speed FiWi broadband access networks, therefore, it is essential to analyze and reduce the typical problems (e.g., bandwidth and delay) in the high-speed next-generation networks (NGANs). This study investigates the performance of a fiber wireless architecture where a 10-Gigabit-capable passive optical network (XGPON) and fifth generation of wireless local area network (WLAN) (i.e., IEEE 802.11ac) are integrated. Both technologies take benefits from each other and have pros and cons concerning the QoS demands of subscribers. The proposed work offers a very flexible QoS scheme for the different types of services of 5G WLAN and XGPON with the help of the highest cost first (HCF) algorithm, which leads to reduced upstream delays for delay-sensitive applications. The simulation results show that the HCF algorithm boosts the performance of the dynamic bandwidth assignment (DBA) scheme and results in up to 96.1%, 90.8%, and 55.5% reduced upstream (US) delays for video: VI(T2), background: BK(T3), and best effort: BE(T4) traffic in enhanced-distributed-channel-access (EDCA) mode. Compared to earlier work, the HCF and immediate allocation with the colorless grant (IACG) DBA combination results in the reduction of up to 54.8% and 53.4% mean US delays. This happens because of 50% to 65% better bandwidth assignment by the IACG DBA process due to efficient mapping by the HCF algorithm.

30 January 2019 in EAI Endorsed Transactions on Creative Technologies
Verified
Journal Article
Open access
PDF file
Rajesh KumarRajesh Kumar
Rajesh Kumar
NSNoman Sarwar
Noman Sarwar
KMKomal Maheshwari
Komal Maheshwari
PLParas Lal
Paras Lal
KDKapeel Dev
Kapeel Dev

29 September 2025 in Cluster Computing
Verified
Journal Article
Open access
PDF file
XZxiao Zheng
xiao Zheng
MTMuhammad Tahir
Muhammad Tahir
JFJaroslav Frnda
Jaroslav Frnda
MAMuhammad Shahid Anwar
Muhammad Shahid Anwar
ACAhyoung Choi
Ahyoung Choi
Rajesh KumarRajesh Kumar
Rajesh Kumar
Mobile edge computing meets stringent latency requirements by offloading computational tasks to edge servers. However, in dynamic and uncertain environments, assigning tasks to multiple users becomes complex. To address this challenge, we design a multi-user task offloading framework that allows users to initiate service requests in a distributed manner. Specifically, we propose an online learning offloading algorithm based on a distributed auction multi-armed bandit, which can adapt to stochastically changing environments and gradually reduce computational latency. We then transform the dynamic task allocation problem into an online multi-user multi-armed bandit problem and develop an offloading algorithm based on heterogeneous distributed multi-armed bandit (HD-MAB) to optimize user rewards subject to network latency. We demonstrate that the HD-MAB algorithm can achieve optimal task allocation, thereby providing near-optimal service performance with linear regret. Simulation results show that our offloading method performs well in optimizing latency-sensitive tasks, and that user participation in the decision-making process of the HD-MAB algorithm does not affect the asymptotic optimality of the algorithm.

24 January 2023
Verified
Conference Paper
Closed access
Rajesh KumarRajesh Kumar
Rajesh Kumar
A lot of research has been done over the last few years in the field of facial Emotion recognition (FER) and it is currently a hot topic in both academia and industry. A good way to perceive a user's thoughts and design a good to detect universal facial expressions system. The option to separate different universal facial expressions based on emotional classes makes it feasible to recognize distinct emotions of facial expressions based on the diversity among facial calculations and expressions. Face recognition technology is used extensively today, often in the medical industry, the attendance system, and so on. This work focuses on developing a Facial Expression Recognition (FER) system using CNN. This the model uses Haar cascade classifier for real-time universal facial expressions using a webcam. It was trained using the [FER2013] dataset available on Kaggle to detect facial expressions.