Satish Kumar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Satish Kumar has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Gopi Ram
National Institute of Technology Warangal
2 shared publications Durbadal Mandal
1 shared publication- Durbadal Mandal
Department of ECE, NIT Durgapur, Durgapur, India
1 shared publication Mridul Kanti Sikder
1 shared publicationRajib Kar
1 shared publication- Rajib Kar
Electronics and Communication Engineering, National Institute of Technology, Durgapur, India
1 shared publication
Latest Publications
- Open Access
- 25 February 2026
- Published by Elsevier in SSRN Electronic Journal
Radar Signal Intelligence for Military Surveillance Using Deep Learning in Airborne Target Identification
With the advent of various types of defensive systems, modern-day surveillance, defense, and airspace security systems require the proper classification of airborne targets using radar for their own application. The threats have grown more complex in detecting and discriminating between hostile flying objects and the benign flying objects, such as birds, with the introduction of small unmanned aerial vehicles (SUAVs), rotorcraft, bionic birds, and similar intruders, as they all share radar micro-Doppler signatures. These signatures are a rich class of motion-related features that, however, often bear close similarity to one another with respect to the different motion classes and thus form a challenge for proper classification. In the present work, we investigate the performance of deep learning and hybrid machine learning algorithms for improving radar-based object recognition. Concretely, the investigation is implemented over three models—CNN, CNN+SVM, and MobileNetV2—over a radar data set with four classes: 2-blade rotor, 3-long-blade rotor, drone, and bird. Both signal (one dimension) and spectrogram image (two dimensions) data were considered. Extensive preprocessing was carried out, and model tuning was performed throughout. Rather than focusing strictly on numerical outputs, the study looks into radar feature extraction from each architecture and adaptation for classification. Performance was evaluated based on accuracy, precision, recall, and F1-score. With 96% overall accuracy, the CNN model does appear to be well suited for real-time classification. This research builds a ground-level approach to integrating intelligent deep learning systems into radar-based airborne target detection systems for better threat identification and response.
- 25 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRadar Signal Intelligence for Military Surveillance Using Deep Learning in Airborne Target Identification
With the advent of various types of defensive systems, modern-day surveillance, defense, and airspace security systems require the proper classification of airborne targets using radar for their own application. The threats have grown more complex in detecting and discriminating between hostile flying objects and the benign flying objects, such as birds, with the introduction of small unmanned aerial vehicles (SUAVs), rotorcraft, bionic birds, and similar intruders, as they all share radar micro-Doppler signatures. These signatures are a rich class of motion-related features that, however, often bear close similarity to one another with respect to the different motion classes and thus form a challenge for proper classification. In the present work, we investigate the performance of deep learning and hybrid machine learning algorithms for improving radar-based object recognition. Concretely, the investigation is implemented over three models—CNN, CNN+SVM, and MobileNetV2—over a radar data set with four classes: 2-blade rotor, 3-long-blade rotor, drone, and bird. Both signal (one dimension) and spectrogram image (two dimensions) data were considered. Extensive preprocessing was carried out, and model tuning was performed throughout. Rather than focusing strictly on numerical outputs, the study looks into radar feature extraction from each architecture and adaptation for classification. Performance was evaluated based on accuracy, precision, recall, and F1-score. With 96% overall accuracy, the CNN model does appear to be well suited for real-time classification. This research builds a ground-level approach to integrating intelligent deep learning systems into radar-based airborne target detection systems for better threat identification and response.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 10, 334-347
Optimal Configuration and Performance Enhancement of Time-Modulated Circular Antenna Arrays
In order to optimize the synthesis of Asymmetric Time-Modulated Circular Antenna Array (ATMCAA) and Symmetric Time-Modulated Circular Antenna Array (STMCAA), this work presents the Novel Particle Swarm Optimization Algorithm (NPSO). Inter-element spacing and uniform current excitation are maintained by regulating the switching time sequence and progressive phase delay of each element. A distinct cost function is developed for each of the two case studies. Using 20- and 36-element examples, several low side-lobe designs synthesized from ATMCAA and STMCAA are compared with traditional circular arrays. Through the manipulation of switching time sequence and progressive phase delay, the cost function is optimized to simultaneously reduce the side-lobe level (SLL) and directivity in ATMCAA and STMCAA. When it comes to antenna array synthesis, NPSO performs better than other algorithms, such as cat swarm optimization and invasive weed optimization. This study demonstrates how effective NPSO is at optimizing antenna arrays in order to improve higher communication reliability and signal quality.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 10, 334-347
Optimal Configuration and Performance Enhancement of Time-Modulated Circular Antenna Arrays
In order to optimize the synthesis of Asymmetric Time-Modulated Circular Antenna Array (ATMCAA) and Symmetric Time-Modulated Circular Antenna Array (STMCAA), this work presents the Novel Particle Swarm Optimization Algorithm (NPSO). Inter-element spacing and uniform current excitation are maintained by regulating the switching time sequence and progressive phase delay of each element. A distinct cost function is developed for each of the two case studies. Using 20- and 36-element examples, several low side-lobe designs synthesized from ATMCAA and STMCAA are compared with traditional circular arrays. Through the manipulation of switching time sequence and progressive phase delay, the cost function is optimized to simultaneously reduce the side-lobe level (SLL) and directivity in ATMCAA and STMCAA. When it comes to antenna array synthesis, NPSO performs better than other algorithms, such as cat swarm optimization and invasive weed optimization. This study demonstrates how effective NPSO is at optimizing antenna arrays in order to improve higher communication reliability and signal quality.