Palash Kumar Kundu
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
Palash Kumar Kundu has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Alok Mukherjee
Electrical Engineering, Government College of Engineering and Ceramic Technology, Kolkata, India
8 shared publications - Arabinda Das
Department of Electrical Engineering, Jadavpur University, Kolkata, India
8 shared publications - Apurba Ghosh
Department of Instrumentation Engineering Burdwan University Burdwan India
6 shared publications - Gautam Sarkar
Department of Electrical Engineering, Jadavpur University, Kolkata, India
5 shared publications - Chayan Kumar Basak
A.P.C. Ray Polytechnic
4 shared publications - Deep Mukherjee
KIIT University
4 shared publications
Latest Publications
- Open Access
- 26 March 2026
- Published by Elsevier in SSRN Electronic Journal
Multi-descriptor hybrid fusion of deep, latent and handcrafted features for reliable thermal tomographic image classification
Thermal tomographic imaging, driven by sensor-based boundary measurements, provides a powerful non-invasive means for reconstructing internal temperature distributions in thermally driven flow systems, enabling predictive maintenance and quality assurance in industrial environments. However, accurate classification of thermal tomographic images remains challenging due to complex heating patterns and the limited availability of annotated datasets. This article presents a multi-descriptor hybrid feature-fusion framework for thermal tomographic image classification, built upon a ResNet-18 deep transfer learning backbone to extract discriminative representations of spatial temperature fields. To enhance robustness, deep features are fused with stacked autoencoder (SAE)-based latent and handcrafted ‘histogram of oriented gradients’ (HOG) descriptors refined through ‘principal component analysis’ (PCA). Tomographic dataset is synthetically constructed using COMSOL Multiphysics with FEM-based simulations of a contour inside a hot airflow chamber, emulating boundary sensor measurements under diverse heating and airflow conditions. The fused-feature set is evaluated using linear ‘support vector machine’ (SVM), ‘logistic regression’ (LR) and ‘linear discriminant analysis’ (LDA) classifiers. Performance is assessed through stratified five-fold cross-validation and holdout validation using overall accuracy, macro F1-score and Cohen’s kappa. Experimental results demonstrate that the proposed configuration consistently outperforms alternative hybrid variants and individual feature branches, achieving classification accuracies up to 99.05% with strong inter-class agreement. Ablation, ROC-AUC, centroid separability and noise-robustness analyses further confirm the compactness and discriminative power of the fused-feature space. The proposed framework offers a reliable solution for automated detection of heating patterns in thermal tomographic imaging, with strong potential for deployment in industrial condition monitoring and non-destructive evaluation systems.
- 26 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-descriptor hybrid fusion of deep, latent and handcrafted features for reliable thermal tomographic image classification
Thermal tomographic imaging, driven by sensor-based boundary measurements, provides a powerful non-invasive means for reconstructing internal temperature distributions in thermally driven flow systems, enabling predictive maintenance and quality assurance in industrial environments. However, accurate classification of thermal tomographic images remains challenging due to complex heating patterns and the limited availability of annotated datasets. This article presents a multi-descriptor hybrid feature-fusion framework for thermal tomographic image classification, built upon a ResNet-18 deep transfer learning backbone to extract discriminative representations of spatial temperature fields. To enhance robustness, deep features are fused with stacked autoencoder (SAE)-based latent and handcrafted ‘histogram of oriented gradients’ (HOG) descriptors refined through ‘principal component analysis’ (PCA). Tomographic dataset is synthetically constructed using COMSOL Multiphysics with FEM-based simulations of a contour inside a hot airflow chamber, emulating boundary sensor measurements under diverse heating and airflow conditions. The fused-feature set is evaluated using linear ‘support vector machine’ (SVM), ‘logistic regression’ (LR) and ‘linear discriminant analysis’ (LDA) classifiers. Performance is assessed through stratified five-fold cross-validation and holdout validation using overall accuracy, macro F1-score and Cohen’s kappa. Experimental results demonstrate that the proposed configuration consistently outperforms alternative hybrid variants and individual feature branches, achieving classification accuracies up to 99.05% with strong inter-class agreement. Ablation, ROC-AUC, centroid separability and noise-robustness analyses further confirm the compactness and discriminative power of the fused-feature space. The proposed framework offers a reliable solution for automated detection of heating patterns in thermal tomographic imaging, with strong potential for deployment in industrial condition monitoring and non-destructive evaluation systems.
- 17 March 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (6)
FEATURE-FUSED STACKED AUTOENCODERS FOR ROBUST THERMAL TOMOGRAPHIC IMAGE CLASSIFICATION
Thermal tomographic imaging is an emerging modality for analyzing internal heat-flow behaviour and identifying localized heating anomalies in complex engineering systems. Classifying such images remains challenging due to their high dimensionality, susceptibility to measurement noise and inadequate availability of annotated datasets. This study develops two lightweight deep feature-learning frameworks based on Stacked Autoencoders (SAE) and a proposed adaptive weighted Multi-Scale Feature-Fused SAE (MSFF-SAE) for robust classification of thermal tomographic images. The SAE captures hierarchical abstractions through progressive compression and reconstruction of input data, supported by sparsity regularization, L2 penalty, a fine-tuned supervised classifier and stochastic noise injection during training. The MSFF-SAE further enhances class separability by fusing multi-level latent features from different encoder depths with the bottleneck representation to preserve both local and global thermal cues. A synthetically generated dataset of 2,070 thermal tomographic images derived from a controlled hot air-flow chamber and COMSOL simulations is used for evaluation. Under a 5-fold cross-validation protocol, adaptive weighted MSFF-SAE achieves a classification accuracy of 97.78 ± 0.93%, outperforming the baseline SAE (96.96 ± 1.45%) with reduced inter-fold variability, indicating stable generalization. Both models exhibit low standard deviation (less than 2%) and macro-averaged ROC-AUC values exceeding 0.99 across folds. The noise-robustness analysis also confirms MSFF-SAE’s superior measurement noise-resilience under increasing Gaussian perturbations. These findings highlight the potential of feature-fused stacked autoencoders as efficient and noise-resilient learners for automated assessment of heating non-uniformities of thermal tomographic images.
- 17 March 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (6)
FEATURE-FUSED STACKED AUTOENCODERS FOR ROBUST THERMAL TOMOGRAPHIC IMAGE CLASSIFICATION
Thermal tomographic imaging is an emerging modality for analyzing internal heat-flow behaviour and identifying localized heating anomalies in complex engineering systems. Classifying such images remains challenging due to their high dimensionality, susceptibility to measurement noise and inadequate availability of annotated datasets. This study develops two lightweight deep feature-learning frameworks based on Stacked Autoencoders (SAE) and a proposed adaptive weighted Multi-Scale Feature-Fused SAE (MSFF-SAE) for robust classification of thermal tomographic images. The SAE captures hierarchical abstractions through progressive compression and reconstruction of input data, supported by sparsity regularization, L2 penalty, a fine-tuned supervised classifier and stochastic noise injection during training. The MSFF-SAE further enhances class separability by fusing multi-level latent features from different encoder depths with the bottleneck representation to preserve both local and global thermal cues. A synthetically generated dataset of 2,070 thermal tomographic images derived from a controlled hot air-flow chamber and COMSOL simulations is used for evaluation. Under a 5-fold cross-validation protocol, adaptive weighted MSFF-SAE achieves a classification accuracy of 97.78 ± 0.93%, outperforming the baseline SAE (96.96 ± 1.45%) with reduced inter-fold variability, indicating stable generalization. Both models exhibit low standard deviation (less than 2%) and macro-averaged ROC-AUC values exceeding 0.99 across folds. The noise-robustness analysis also confirms MSFF-SAE’s superior measurement noise-resilience under increasing Gaussian perturbations. These findings highlight the potential of feature-fused stacked autoencoders as efficient and noise-resilient learners for automated assessment of heating non-uniformities of thermal tomographic images.
- 27 February 2026
- Published by Springer Nature in Cardiovascular Engineering and Technology
- Vol. 17 (2) , 155-174
Deep Learning Based Approach for Lossless ECG Compression
Objective Tele-monitoring is a useful platform for remote monitoring of cardiac patients, where compression plays a significant role in reducing the link burden and memory utilization of the source device. This paper describes a new approach for lossless ECG compression based on a deep-learning method via an adaptive autoregressive integrated moving average (ARIMA) model. Methods Raw ECG signals were denoised and preprocessed to generate beat-cells for further processing. The ARIMA model uses the individual cardiac cycles to generate model parameters, which are then compressed. In this research, the optimal model hyperparameters were predicted by a deep autoencoder followed by a multilayer perceptron neural network (MLPNN) regressor combination. The predictor was tuned offline via particle swarm optimization (PSO), which produced the reference data for MLPNN tuning. Results The technique uses 46 records of mitdb under PhysioNet, including 10 major abnormal beats: H, A, V, P, L, R, a, f, F and j. Because of the adaptive nature, compression quality is high with negligible loss. No deviations in the clinical features of the reconstructed beats are found. The mean CR and PRD% values were 41.51 and 0.209%, respectively, which are superior to those reported in published research on ECG compression. Conclusion The proposed adaptive ECG compression model can be useful for real-time telemonitoring applications, efficient storage and transmission of streamlined data of critical patients under continuous monitoring.
- 27 February 2026
- Published by Springer Nature in Cardiovascular Engineering and Technology
- Vol. 17 (2) , 155-174
Deep Learning Based Approach for Lossless ECG Compression
Objective Tele-monitoring is a useful platform for remote monitoring of cardiac patients, where compression plays a significant role in reducing the link burden and memory utilization of the source device. This paper describes a new approach for lossless ECG compression based on a deep-learning method via an adaptive autoregressive integrated moving average (ARIMA) model. Methods Raw ECG signals were denoised and preprocessed to generate beat-cells for further processing. The ARIMA model uses the individual cardiac cycles to generate model parameters, which are then compressed. In this research, the optimal model hyperparameters were predicted by a deep autoencoder followed by a multilayer perceptron neural network (MLPNN) regressor combination. The predictor was tuned offline via particle swarm optimization (PSO), which produced the reference data for MLPNN tuning. Results The technique uses 46 records of mitdb under PhysioNet, including 10 major abnormal beats: H, A, V, P, L, R, a, f, F and j. Because of the adaptive nature, compression quality is high with negligible loss. No deviations in the clinical features of the reconstructed beats are found. The mean CR and PRD% values were 41.51 and 0.209%, respectively, which are superior to those reported in published research on ECG compression. Conclusion The proposed adaptive ECG compression model can be useful for real-time telemonitoring applications, efficient storage and transmission of streamlined data of critical patients under continuous monitoring.
- Open Access
- 16 January 2026
- Published by Springer Nature in Multiscale and Multidisciplinary Modeling, Experiments and Design
- Vol. 9 (1) , 77
Deep transfer learning framework for thermal tomographic image classification: a comparative study
- 16 January 2026
- Published by Springer Nature in Multiscale and Multidisciplinary Modeling, Experiments and Design
- Vol. 9 (1) , 77
Open AccessDeep transfer learning framework for thermal tomographic image classification: a comparative study
- 9 July 2025
- Published by Wiley in International Journal of Robust and Nonlinear Control
- Vol. 35 (17) , 7422-7437
Fractional Standalone Backstepping Control for Nonlinear Continuous Stirred Tank Reactor
Controlling the molar concentration in a chemical reactor is a challenging task in the presence of noise and disturbance. Contrary to the reported works based on augmented method, this work proposes a novel fractional order standalone backstepping rule as a feasible alternative to stabilize the concentration of a nonlinear continuous stirred tank reactor process. The novel indirect biquadratic equiripple approximation technique is suggested to realize the behavior of fractional order element in the proposed rule with a suitable integer order process. The proposed approximation method is further modified by an additional exact phase approach to enhance the behavior of fractional order element. A comprehensive simulation study on both frequency and time domain platforms is carried out to establish the suitable fractional order in the proposed law. Closed‐loop performance and control efforts obtained through simulation studies vindicate that the proposed technique outperforms existing control strategies without unnecessary overshoots and yields relatively smooth control action in the presence of noise and disturbance. By abruptly altering the system settings, the robust stability investigation of the proposed rule is also carried out in the presence of load disturbance and noise. This proposed technique offers greater accuracy by resulting in less integral errors and total variation of input control action in both nominal and perturbed circumstances.
- 9 July 2025
- Published by Wiley in International Journal of Robust and Nonlinear Control
- Vol. 35 (17) , 7422-7437
Fractional Standalone Backstepping Control for Nonlinear Continuous Stirred Tank Reactor
Controlling the molar concentration in a chemical reactor is a challenging task in the presence of noise and disturbance. Contrary to the reported works based on augmented method, this work proposes a novel fractional order standalone backstepping rule as a feasible alternative to stabilize the concentration of a nonlinear continuous stirred tank reactor process. The novel indirect biquadratic equiripple approximation technique is suggested to realize the behavior of fractional order element in the proposed rule with a suitable integer order process. The proposed approximation method is further modified by an additional exact phase approach to enhance the behavior of fractional order element. A comprehensive simulation study on both frequency and time domain platforms is carried out to establish the suitable fractional order in the proposed law. Closed‐loop performance and control efforts obtained through simulation studies vindicate that the proposed technique outperforms existing control strategies without unnecessary overshoots and yields relatively smooth control action in the presence of noise and disturbance. By abruptly altering the system settings, the robust stability investigation of the proposed rule is also carried out in the presence of load disturbance and noise. This proposed technique offers greater accuracy by resulting in less integral errors and total variation of input control action in both nominal and perturbed circumstances.
- 28 May 2025
- Published by Taylor & Francis in LEUKOS
- Vol. 22 (3) , 316-325
Real-Time Implementation of Cloud-Based Indoor Illumination Monitoring System
- 28 May 2025
- Published by Taylor & Francis in LEUKOS
- Vol. 22 (3) , 316-325
Real-Time Implementation of Cloud-Based Indoor Illumination Monitoring System
- 13 March 2025
- Published by Springer Nature in Sensing and Imaging
- Vol. 26 (1) , 1-19
Assessing Heating Quality Using Multivariate Statistical Techniques for Thermal Tomographic Images
- 13 March 2025
- Published by Springer Nature in Sensing and Imaging
- Vol. 26 (1) , 1-19
Assessing Heating Quality Using Multivariate Statistical Techniques for Thermal Tomographic Images
- 28 January 2024
- Published by Springer Nature in Sādhanā
- Vol. 49 (1) , 1-16
Analysis of improved fractional backstepping and lyapunov strategies for stabilization of inverted pendulum
- 28 January 2024
- Published by Springer Nature in Sādhanā
- Vol. 49 (1) , 1-16
Analysis of improved fractional backstepping and lyapunov strategies for stabilization of inverted pendulum
- Open Access
- 4 January 2024
- Published by Association for Computing Machinery (ACM)
- p. 588-589
CNN and MLR based Multifunction Sensor Development using Piezo-resistive Element
In the present work, a multifunction sensor is developed using convolutional neural network (CNN) and multiple linear regression to measure temperature and pressure simultaneously. First, the hardware sensor is built using piezo-resistive material. When the hardware sensor is excited by electric signals, it produces an aggregated output (emf). A hybrid CNN and MLR based approach is devised to extract temperature and pressure values from the aggregated output.
- 4 January 2024
- Published by Association for Computing Machinery (ACM)
- p. 588-589
Open AccessCNN and MLR based Multifunction Sensor Development using Piezo-resistive Element
In the present work, a multifunction sensor is developed using convolutional neural network (CNN) and multiple linear regression to measure temperature and pressure simultaneously. First, the hardware sensor is built using piezo-resistive material. When the hardware sensor is excited by electric signals, it produces an aggregated output (emf). A hybrid CNN and MLR based approach is devised to extract temperature and pressure values from the aggregated output.