Narendra Digambar Londhe
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
Narendra Digambar Londhe has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Ghanahshyam B. Kshirsagar
College of Engineering, The Roux Institute, Northeastern University, Portland, ME, USA
9 shared publications - Rajendra S. Sonawane
Psoriasis Clinic and Research Centre, Psoriatreat, Pune, Maharashtra, India
8 shared publications - Jasjit S. Suri
Department of ECE, Idaho State University, Pocatello, ID 83209, USA
6 shared publications - Ritesh Raj
Amrita Vishwa Vidyapeetham, Karnataka, India
5 shared publications - Vimal Kumar Shrivastava
School of Electronics Engineering, Kalinga Institute of Industrial Technology (KIIT), Bhubaneswar, India
5 shared publications - Vibha Bhandari
Electrical Engineering Department, National Institute of Technology, Raipur, Chhattisgarh, India
4 shared publications
Latest Publications
- 7 September 2026
- Published by Elsevier in Computers and Electrical Engineering
- Vol. 140
A novel self-supervised deep learning model for multi-class fault classification in transmission lines
- 7 September 2026
- Published by Elsevier in Computers and Electrical Engineering
- Vol. 140
A novel self-supervised deep learning model for multi-class fault classification in transmission lines
- 1 May 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (10) , 105324
Deep learning transformer with extreme gradient boosting for precise fault distance estimation in power transmission system
Fault distance estimation in multi-bus transmission lines is challenging but essential for reducing outage time and labor costs. Conventional fault distance estimation schemes face challenges, such as the impedance-based methods necessitating comprehensive knowledge about the line parameters, and the traveling wave-based method requires a high sampling frequency for fault distance estimation. While using artificial intelligence, neither the fault resistance influences the fault location, nor is any information about the source impedance needed. Machine learning requires manual feature extraction and data preprocessing, which adds extra work. This paper employs a hybrid deep learning model based on Transformer and extreme gradient boosting (XGBoost), where the Transformer extracts the prominent feature from raw fault data and XGBoost predicts the fault distance in power transmission systems. The proposed model achieved a mean absolute error (MAE) ranging from 0.0008 to 0.6406 across different lines and an average MAE of 0.27025 km for all lines using an 80:20 train/test split, demonstrating strong performance. Ablation studies further validated the model’s robustness against noise, bad data, sensitivity to line parameter variation, load variation and different train/test splits. For validation, the data is generated from a real-time simulator, and unseen fault data is provided to a Raspberry Pi microcontroller deployed with the proposed model for fault distance estimation.
- 1 May 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (10) , 105324
Deep learning transformer with extreme gradient boosting for precise fault distance estimation in power transmission system
Fault distance estimation in multi-bus transmission lines is challenging but essential for reducing outage time and labor costs. Conventional fault distance estimation schemes face challenges, such as the impedance-based methods necessitating comprehensive knowledge about the line parameters, and the traveling wave-based method requires a high sampling frequency for fault distance estimation. While using artificial intelligence, neither the fault resistance influences the fault location, nor is any information about the source impedance needed. Machine learning requires manual feature extraction and data preprocessing, which adds extra work. This paper employs a hybrid deep learning model based on Transformer and extreme gradient boosting (XGBoost), where the Transformer extracts the prominent feature from raw fault data and XGBoost predicts the fault distance in power transmission systems. The proposed model achieved a mean absolute error (MAE) ranging from 0.0008 to 0.6406 across different lines and an average MAE of 0.27025 km for all lines using an 80:20 train/test split, demonstrating strong performance. Ablation studies further validated the model’s robustness against noise, bad data, sensitivity to line parameter variation, load variation and different train/test splits. For validation, the data is generated from a real-time simulator, and unseen fault data is provided to a Raspberry Pi microcontroller deployed with the proposed model for fault distance estimation.
- 2 January 2026
- Published by Springer Nature
- p. 189-197
Weekly Demand Load Forecasting Using LSTM Model for Chhattisgarh State
- 2 January 2026
- Published by Springer Nature
- p. 189-197
Weekly Demand Load Forecasting Using LSTM Model for Chhattisgarh State
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 56 (1) , 124-134
P3-FSLNet: A Compact Spatio-Temporal Model With Contrastive Few-Shot Learning for Subject-Independent P300 Detection in Devanagari Script-Based P300 Speller
Brain–computer interface (BCI) systems frequently necessitate time-intensive subject-specific calibration, thereby motivating the development of subject-independent P300 detection approaches. Existing methodologies that employ transfer learning and knowledge distillation encounter challenges with limited generalizability due to substantial intersubject variability and constrained data availability. Furthermore, their applicability is often questioned, as most lack comprehensive external validation and cross-script evaluation. To address these limitations, we introduce P3-FSLNet, a few-shot metalearning framework that integrates prototypical networks and contrastive learning within a spatial-temporal convolutional neural network augmented by dual-channel attention for the selection of relevant electroencephalogram channels. Episodic metatraining facilitates the transfer of robust knowledge across different subjects. Evaluated on a self-recorded Devanagari script dataset, P3-FSLNet attains a classification accuracy of 93.17%, surpassing state-of-the-art methods by 1%–14%, while simultaneously reducing trainable parameters by up to 400 times. External validation using English-script datasets from BCI Competition II and III confirms its robustness in cross-subject and cross-script generalization. These findings demonstrate the efficacy of P3-FSLNet and represent a substantial advancement toward the development of script-agnostic P300 spellers that are lightweight, scalable, and conducive to multilingual applications.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 56 (1) , 124-134
P3-FSLNet: A Compact Spatio-Temporal Model With Contrastive Few-Shot Learning for Subject-Independent P300 Detection in Devanagari Script-Based P300 Speller
Brain–computer interface (BCI) systems frequently necessitate time-intensive subject-specific calibration, thereby motivating the development of subject-independent P300 detection approaches. Existing methodologies that employ transfer learning and knowledge distillation encounter challenges with limited generalizability due to substantial intersubject variability and constrained data availability. Furthermore, their applicability is often questioned, as most lack comprehensive external validation and cross-script evaluation. To address these limitations, we introduce P3-FSLNet, a few-shot metalearning framework that integrates prototypical networks and contrastive learning within a spatial-temporal convolutional neural network augmented by dual-channel attention for the selection of relevant electroencephalogram channels. Episodic metatraining facilitates the transfer of robust knowledge across different subjects. Evaluated on a self-recorded Devanagari script dataset, P3-FSLNet attains a classification accuracy of 93.17%, surpassing state-of-the-art methods by 1%–14%, while simultaneously reducing trainable parameters by up to 400 times. External validation using English-script datasets from BCI Competition II and III confirms its robustness in cross-subject and cross-script generalization. These findings demonstrate the efficacy of P3-FSLNet and represent a substantial advancement toward the development of script-agnostic P300 spellers that are lightweight, scalable, and conducive to multilingual applications.
- 11 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 3420-3424
A Technical Odyssey of Self-Supervised Representation Learning for Devanagari-Script-Based P300 Speller
Traditional supervised learning (SL) methods for P300 event-related potential (ERP) detection in P300 spellers require extensive labelled data and often struggle to generalize well across subjects and trials, especially with limited data. Previous efforts using transfer learning and knowledge distillation improved performance but still face high computational complexity and lack transparency. These issues highlight the need to explore new approaches to enhance transferability and reduce uncertainty. To address this, we investigated the effectiveness of representational learning through a self-supervised approach. Our self-supervised learning (SSL) framework, featuring a compact convolutional neural network (CNN) backbone and label-agnostic characteristics, improves the robustness of learned features to variations in ERPs encountered in P300 speller. Experiments on self-recorded data and ablation studies show that the learned representations are robust and effective. Achieving an accuracy of 84%, the downstream classifier trained on the SSL framework performed competitively with traditional supervised methods. Additionally, comparison between features learned with SL and SSL, using t-SNE visualization and correlation coefficient (r = -0.51) analysis, demonstrates that SSL features offer better discrimination between P300 and non-P300.
- 11 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 3420-3424
A Technical Odyssey of Self-Supervised Representation Learning for Devanagari-Script-Based P300 Speller
Traditional supervised learning (SL) methods for P300 event-related potential (ERP) detection in P300 spellers require extensive labelled data and often struggle to generalize well across subjects and trials, especially with limited data. Previous efforts using transfer learning and knowledge distillation improved performance but still face high computational complexity and lack transparency. These issues highlight the need to explore new approaches to enhance transferability and reduce uncertainty. To address this, we investigated the effectiveness of representational learning through a self-supervised approach. Our self-supervised learning (SSL) framework, featuring a compact convolutional neural network (CNN) backbone and label-agnostic characteristics, improves the robustness of learned features to variations in ERPs encountered in P300 speller. Experiments on self-recorded data and ablation studies show that the learned representations are robust and effective. Achieving an accuracy of 84%, the downstream classifier trained on the SSL framework performed competitively with traditional supervised methods. Additionally, comparison between features learned with SL and SSL, using t-SNE visualization and correlation coefficient (r = -0.51) analysis, demonstrates that SSL features offer better discrimination between P300 and non-P300.
- Open Access
- 10 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 51-56
Transfer Learning Approach for Assessment of Psoriasis Erythema Severity
- 10 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 51-56
Open AccessTransfer Learning Approach for Assessment of Psoriasis Erythema Severity
- 4 December 2024
- Published by Springer Nature
- p. 145-159
Overcomplete U-Net Networks for Psoriasis Detection in Digital Color Images
- 4 December 2024
- Published by Springer Nature
- p. 145-159
Overcomplete U-Net Networks for Psoriasis Detection in Digital Color Images
- 11 August 2024
- Published by Springer Nature in Journal of Control, Automation and Electrical Systems
- Vol. 35 (5) , 884-898
Fault-MTL: A Multi-task Deep Learning Approach for Simultaneous Fault Classification and Localization in Power Systems
- 11 August 2024
- Published by Springer Nature in Journal of Control, Automation and Electrical Systems
- Vol. 35 (5) , 884-898
Fault-MTL: A Multi-task Deep Learning Approach for Simultaneous Fault Classification and Localization in Power Systems
- Open Access
- 5 March 2024
- Published by BON VIEW PUBLISHING PTE in Artificial Intelligence and Applications
- Vol. 2 (3) , 155-164
A Systematic Review of Computational Intelligence Techniques for Channel Selection in P300-Based Brain Computer Interface Speller
Electroencephalography (EEG)-based P300 speller aids in restoring the communication and control capabilities in patients suffering from motor disabilities. However, the quality and quantity of the data collected from EEG recordings have a substantial influence on the P300 speller’s performance. Hence, selecting the optimum number of recording electrodes, i.e., channels for each user, is a significant difficulty for the P300 speller. There are two fundamental objectives of the channel selection process: (1) to extract the most crucial information from the relevant channels, hence reducing the computing complexity of P300/non-P300 signal processing operation, and (2) to lessen the potential overfitting that could result from using unwanted channels to boost performance. For obtaining the best channel subsets, different channel selection techniques, including manual, filtering, wrapper, and embedded approaches, have been applied by past researchers. This research provides an in-depth examination of recent advancements, status, challenges, and potential solutions related to channel selection strategies in P300 speller systems. Each channel selection technique is thoroughly explored, including detailed comparisons between them. The notable advantages and drawbacks of each method are emphasized along with the discussion on the future direction and scope of work in the field of channel selection in P300 speller. The review underscores that channel selection methods enable the use of a reduced number of channels without compromising classification performance. By eliminating noisy or irrelevant channels, these approaches contribute to enhanced system performance. Received: 21 July 2023 | Revised: 3 January 2024 | Accepted: 29 February 2024 Conflicts of Interest Narendra D. Londhe is an Editorial Board Member for Artificial Intelligence and Applications, and was not involved in the editorial review or the decision to publish this article. The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study.
- 5 March 2024
- Published by BON VIEW PUBLISHING PTE in Artificial Intelligence and Applications
- Vol. 2 (3) , 155-164
Open AccessA Systematic Review of Computational Intelligence Techniques for Channel Selection in P300-Based Brain Computer Interface Speller
Electroencephalography (EEG)-based P300 speller aids in restoring the communication and control capabilities in patients suffering from motor disabilities. However, the quality and quantity of the data collected from EEG recordings have a substantial influence on the P300 speller’s performance. Hence, selecting the optimum number of recording electrodes, i.e., channels for each user, is a significant difficulty for the P300 speller. There are two fundamental objectives of the channel selection process: (1) to extract the most crucial information from the relevant channels, hence reducing the computing complexity of P300/non-P300 signal processing operation, and (2) to lessen the potential overfitting that could result from using unwanted channels to boost performance. For obtaining the best channel subsets, different channel selection techniques, including manual, filtering, wrapper, and embedded approaches, have been applied by past researchers. This research provides an in-depth examination of recent advancements, status, challenges, and potential solutions related to channel selection strategies in P300 speller systems. Each channel selection technique is thoroughly explored, including detailed comparisons between them. The notable advantages and drawbacks of each method are emphasized along with the discussion on the future direction and scope of work in the field of channel selection in P300 speller. The review underscores that channel selection methods enable the use of a reduced number of channels without compromising classification performance. By eliminating noisy or irrelevant channels, these approaches contribute to enhanced system performance. Received: 21 July 2023 | Revised: 3 January 2024 | Accepted: 29 February 2024 Conflicts of Interest Narendra D. Londhe is an Editorial Board Member for Artificial Intelligence and Applications, and was not involved in the editorial review or the decision to publish this article. The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study.