Vijay Bhaskar Semwal
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
Vijay Bhaskar Semwal has a total of 120 co-authors. Here are the top co-authors by number of shared publications:
- Pratibha Tokas
School of Computer Science, UPES, Dehradun, India
11 shared publications - Vishwanath Bijalwan
School of Computer Science and Artificial Intelligence, SR University, Warangal, Telangana, India
7 shared publications G. C. Nandi
Indian Institute of Information Technology, Allahabad
5 shared publicationsG.C. Nandi
5 shared publicationsGora Chand Nandi
Indian Institute of Information Technology, Allahabad, India
4 shared publications- Manish Raj
Bennett University
4 shared publications
Latest Publications
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (4) , 1-4
ZMP Estimation From Wearable Sensor Using Deep Learning for Gait Analysis
Zero moment point (ZMP) is a key biomechanical indicator of dynamic gait stability, conventionally measured using force plates or pressure insoles that are costly and impractical for continuous wearable use. This work presents a low-cost wearable framework for ZMP estimation using only thigh–shank inertial measurement units (IMUs) and knee joint kinematics. A custom flex sensor-based insole was used solely to compute reference ZMP during training and evaluation. Data were collected from healthy subjects walking on flat and inclined treadmills. A bidirectional long short-term memory–artificial neural network model trained on synchronized IMU and knee angle data achieved subjectwise fivefold cross-validation RMSEs of 0.56 cm (mediolateral) and 0.09 cm (anterior–posterior), with corresponding $R^{2}$ values of 0.82 and 0.90. Leave-one-subject-out evaluation demonstrated subject-independent generalization with a modest increase in error due to intersubject variability. Real-time deployment on an NVIDIA Jetson Nano achieved an average inference time of 9.2 ms with a quantized model size of 1.8 MB, enabling on-device ZMP estimation in controlled gait settings.
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (4) , 1-4
ZMP Estimation From Wearable Sensor Using Deep Learning for Gait Analysis
Zero moment point (ZMP) is a key biomechanical indicator of dynamic gait stability, conventionally measured using force plates or pressure insoles that are costly and impractical for continuous wearable use. This work presents a low-cost wearable framework for ZMP estimation using only thigh–shank inertial measurement units (IMUs) and knee joint kinematics. A custom flex sensor-based insole was used solely to compute reference ZMP during training and evaluation. Data were collected from healthy subjects walking on flat and inclined treadmills. A bidirectional long short-term memory–artificial neural network model trained on synchronized IMU and knee angle data achieved subjectwise fivefold cross-validation RMSEs of 0.56 cm (mediolateral) and 0.09 cm (anterior–posterior), with corresponding $R^{2}$ values of 0.82 and 0.90. Leave-one-subject-out evaluation demonstrated subject-independent generalization with a modest increase in error due to intersubject variability. Real-time deployment on an NVIDIA Jetson Nano achieved an average inference time of 9.2 ms with a quantized model size of 1.8 MB, enabling on-device ZMP estimation in controlled gait settings.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 1660-1671
MSA-TCN: An Ultra-Compact, Energy-Efficient Edge TCN for Explainable HAR in Consumer IoHT Devices
Human Activity Recognition (HAR) in consumer healthcare and Internet of Health Things (IoHT) devices demands models that are accurate, energy-efficient, and explainable for on-device deployment. We propose MSA-TCN, an ultra-compact Adaptive Temporal Convolutional Network that combines multi-scale temporal convolutions, causal dilated encoding, and hybrid channel–temporal attention to model both short- and long-range motion dependencies. A subject-aware augmentation strategy enhances generalization across users and sensor configurations without additional computation. The quantized INT8 variant achieves an average accuracy of 98.7% across five benchmark datasets, with a compact size of 0.08 MB and an inference latency of 1.8 ms per 256-sample window on a mid-range smartphone (Snapdragon 845). Energy consumption is reduced by over 98% (from 32.5 mJ to 0.36 mJ) compared to FP32 inference. Post-hoc SHAP analysis provides clinically meaningful attributions, improving transparency and user trust in healthcare settings. By enabling on-device prediction, MSA-TCN minimizes wireless transmission and ensures reliability under intermittent connectivity, offering an interpretable, energy-efficient solution for next-generation IoHT applications.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 1660-1671
MSA-TCN: An Ultra-Compact, Energy-Efficient Edge TCN for Explainable HAR in Consumer IoHT Devices
Human Activity Recognition (HAR) in consumer healthcare and Internet of Health Things (IoHT) devices demands models that are accurate, energy-efficient, and explainable for on-device deployment. We propose MSA-TCN, an ultra-compact Adaptive Temporal Convolutional Network that combines multi-scale temporal convolutions, causal dilated encoding, and hybrid channel–temporal attention to model both short- and long-range motion dependencies. A subject-aware augmentation strategy enhances generalization across users and sensor configurations without additional computation. The quantized INT8 variant achieves an average accuracy of 98.7% across five benchmark datasets, with a compact size of 0.08 MB and an inference latency of 1.8 ms per 256-sample window on a mid-range smartphone (Snapdragon 845). Energy consumption is reduced by over 98% (from 32.5 mJ to 0.36 mJ) compared to FP32 inference. Post-hoc SHAP analysis provides clinically meaningful attributions, improving transparency and user trust in healthcare settings. By enabling on-device prediction, MSA-TCN minimizes wireless transmission and ensures reliability under intermittent connectivity, offering an interpretable, energy-efficient solution for next-generation IoHT applications.
- 2 January 2026
- Published by Springer Nature
- p. 29-36
Manipulator Robot Development and Analysis of Motion Planning Libraries Through ROS2 and MoveIt2
- 2 January 2026
- Published by Springer Nature
- p. 29-36
Manipulator Robot Development and Analysis of Motion Planning Libraries Through ROS2 and MoveIt2
- Open Access
- 15 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Quantum-Resistant Artificial Intelligence Algorithms for Secure Networks
- 15 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessQuantum-Resistant Artificial Intelligence Algorithms for Secure Networks
- 10 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (24) , 44740-44750
An Efficient Deep Learning Framework for Lower Limb Activity Recognition From Wearable Sensors in Knee Pathologies
Human activity recognition (HAR) using wearable sensor signals has gained significant attention for its applications in healthcare, rehabilitation, and personalized monitoring. However, recognizing lower limb activities in individuals with musculoskeletal pathologies remains a challenging task due to signal variability and noise. This study proposes an efficient deep learning framework that integrates advanced signal enhancement and a modular multibranch convolutional architecture for robust activity classification in knee pathologies. A publicly available multichannel surface electromyography (sEMG) dataset was employed, where signals were processed through a hybrid denoising pipeline combining bandpass filtering, wavelet denoising (WD), and ensemble empirical mode decomposition (EEMD). Selective reconstruction of the most informative intrinsic mode functions (IMFs) was performed based on entropy, signal-to-noise ratio (SNR), energy, and correlation metrics. To address class imbalance, adaptive synthetic (ADASYN) sampling was applied, followed by segmentation into 256 ms windows with 25% overlap. The proposed model achieved superior classification performance, attaining average accuracies of 99.8% for healthy subjects and 99.3% for pathological subjects across sitting, standing, and walking activities. The proposed framework was also evaluated on the mHealth dataset with inertial measurement unit (IMU) signals and achieved an average accuracy of 99.3%. Furthermore, to assess its generalization on a more diverse and contemporary benchmark, the model was validated on the WISDM dataset, achieving an average accuracy of 98.6% and an F1-score of 98.4%, demonstrating strong consistency. Despite its high accuracy, the model is compact (27 K parameters, <0.5-MB memory) and real-time deployable, with an average inference time of 15.2 ms per sample. These results demonstrate the effectiveness of the framework in handling noisy and imbalanced wearable sensor data and its potential for clinical decision support and rehabilitation applications.
- 10 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (24) , 44740-44750
An Efficient Deep Learning Framework for Lower Limb Activity Recognition From Wearable Sensors in Knee Pathologies
Human activity recognition (HAR) using wearable sensor signals has gained significant attention for its applications in healthcare, rehabilitation, and personalized monitoring. However, recognizing lower limb activities in individuals with musculoskeletal pathologies remains a challenging task due to signal variability and noise. This study proposes an efficient deep learning framework that integrates advanced signal enhancement and a modular multibranch convolutional architecture for robust activity classification in knee pathologies. A publicly available multichannel surface electromyography (sEMG) dataset was employed, where signals were processed through a hybrid denoising pipeline combining bandpass filtering, wavelet denoising (WD), and ensemble empirical mode decomposition (EEMD). Selective reconstruction of the most informative intrinsic mode functions (IMFs) was performed based on entropy, signal-to-noise ratio (SNR), energy, and correlation metrics. To address class imbalance, adaptive synthetic (ADASYN) sampling was applied, followed by segmentation into 256 ms windows with 25% overlap. The proposed model achieved superior classification performance, attaining average accuracies of 99.8% for healthy subjects and 99.3% for pathological subjects across sitting, standing, and walking activities. The proposed framework was also evaluated on the mHealth dataset with inertial measurement unit (IMU) signals and achieved an average accuracy of 99.3%. Furthermore, to assess its generalization on a more diverse and contemporary benchmark, the model was validated on the WISDM dataset, achieving an average accuracy of 98.6% and an F1-score of 98.4%, demonstrating strong consistency. Despite its high accuracy, the model is compact (27 K parameters, <0.5-MB memory) and real-time deployable, with an average inference time of 15.2 ms per sample. These results demonstrate the effectiveness of the framework in handling noisy and imbalanced wearable sensor data and its potential for clinical decision support and rehabilitation applications.
- 8 November 2025
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 51 (5) , 6709-6722
Adaptive VMD-Based Explainable Hybrid Deep Learning Model for Robust Lower Limb Activity Recognition from sEMG Signals
- 8 November 2025
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 51 (5) , 6709-6722
Adaptive VMD-Based Explainable Hybrid Deep Learning Model for Robust Lower Limb Activity Recognition from sEMG Signals
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (21) , 40461-40471
A Real-Time Deployable Attention-Driven CNN–LSTM Framework for Human Activity Recognition Using Wearable Sensor
Human Activity Recognition (HAR) using wearable inertial sensors is increasingly vital for healthcare, rehabilitation, and pervasive computing. Yet, many state-of-the-art models rely on handcrafted feature engineering, heavy time–frequency transformations, or large neural networks, limiting real-time deployment on resource-constrained smartphones and wearables. We propose an end-to-end, lightweight attention-guided CNN–CBAM–LSTM architecture that directly consumes raw IMU signals standardized by StandardScaler, eliminating costly preprocessing. Our design uniquely combines: (i) multi-resolution convolutional kernels to capture diverse motion dynamics; (ii) Convolutional Block Attention Modules (CBAM) to adaptively highlight salient spatial features and suppress sensor noise; and (iii) temporal self-attention atop stacked LSTM layers to model long-range temporal dependencies. Extensive experiments on three public HAR benchmarks—mHealth (controlled), MotionSense (uncontrolled), and UCI HAR (semi-controlled)—show that our model achieves accuracies of 99.7%, 99.5%, and 99.4% respectively, consistently surpassing recent baselines. Moreover, structured pruning and quantization compress the model to just 130.8 KB, while real-world tests on a Snapdragon 845-class Android device confirm ultra-low inference latency of 13.5 ms per segment under live streaming. This fusion of high accuracy, low latency, and minimal memory footprint demonstrates the potential of attention-driven, multi-resolution deep learning for practical, real-time edge AI in HAR applications.
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (21) , 40461-40471
A Real-Time Deployable Attention-Driven CNN–LSTM Framework for Human Activity Recognition Using Wearable Sensor
Human Activity Recognition (HAR) using wearable inertial sensors is increasingly vital for healthcare, rehabilitation, and pervasive computing. Yet, many state-of-the-art models rely on handcrafted feature engineering, heavy time–frequency transformations, or large neural networks, limiting real-time deployment on resource-constrained smartphones and wearables. We propose an end-to-end, lightweight attention-guided CNN–CBAM–LSTM architecture that directly consumes raw IMU signals standardized by StandardScaler, eliminating costly preprocessing. Our design uniquely combines: (i) multi-resolution convolutional kernels to capture diverse motion dynamics; (ii) Convolutional Block Attention Modules (CBAM) to adaptively highlight salient spatial features and suppress sensor noise; and (iii) temporal self-attention atop stacked LSTM layers to model long-range temporal dependencies. Extensive experiments on three public HAR benchmarks—mHealth (controlled), MotionSense (uncontrolled), and UCI HAR (semi-controlled)—show that our model achieves accuracies of 99.7%, 99.5%, and 99.4% respectively, consistently surpassing recent baselines. Moreover, structured pruning and quantization compress the model to just 130.8 KB, while real-world tests on a Snapdragon 845-class Android device confirm ultra-low inference latency of 13.5 ms per segment under live streaming. This fusion of high accuracy, low latency, and minimal memory footprint demonstrates the potential of attention-driven, multi-resolution deep learning for practical, real-time edge AI in HAR applications.
- Top Cited Papers
- 30 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (12) , 22618-22628
A Lightweight and Explainable Hybrid Deep Learning Model for Wearable Sensor-Based Human Activity Recognition
Human Activity Recognition (HAR) is critical for rehabilitation and clinical monitoring, but robust recognition using wearable sensors (e.g., sEMG or IMU) remains challenging due to signal noise and variability. We propose X-LiteHAR, a lightweight, explainable hybrid deep learning framework for real-time HAR, combining adaptive EEMD for noise-robust signal enhancement and a Multi-Head CNN-LSTM for spatio-temporal feature learning. The optimized framework demonstrates efficient edge deployment through structured pruning and quantization, achieving 70% model size reduction while maintaining competitive performance, with on-device validation on an android OnePlus 6T smartphone showing 9 ms inference latency. The model was trained and evaluated independently on two distinct datasets: (1) the UCI sEMG dataset (muscle activity signals) and (2) the IMU-only MHealth dataset (motion signals), demonstrating the architecture’s adaptability to different sensor modalities. On the UCI dataset, X-LiteHAR achieved 99.0% accuracy (healthy subjects) and 98.7% (pathological), while on MHealth (IMU-only), it reached 99.2% accuracy. Leveraging Explainable AI (XAI), we interpret muscle activation patterns for personalized rehabilitation insights. By unifying signal processing, efficient deep learning, and interpretability, X-LiteHAR advances real-time HAR for clinical and wearable applications.
- 30 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (12) , 22618-22628
Top Cited PapersA Lightweight and Explainable Hybrid Deep Learning Model for Wearable Sensor-Based Human Activity Recognition
Human Activity Recognition (HAR) is critical for rehabilitation and clinical monitoring, but robust recognition using wearable sensors (e.g., sEMG or IMU) remains challenging due to signal noise and variability. We propose X-LiteHAR, a lightweight, explainable hybrid deep learning framework for real-time HAR, combining adaptive EEMD for noise-robust signal enhancement and a Multi-Head CNN-LSTM for spatio-temporal feature learning. The optimized framework demonstrates efficient edge deployment through structured pruning and quantization, achieving 70% model size reduction while maintaining competitive performance, with on-device validation on an android OnePlus 6T smartphone showing 9 ms inference latency. The model was trained and evaluated independently on two distinct datasets: (1) the UCI sEMG dataset (muscle activity signals) and (2) the IMU-only MHealth dataset (motion signals), demonstrating the architecture’s adaptability to different sensor modalities. On the UCI dataset, X-LiteHAR achieved 99.0% accuracy (healthy subjects) and 98.7% (pathological), while on MHealth (IMU-only), it reached 99.2% accuracy. Leveraging Explainable AI (XAI), we interpret muscle activation patterns for personalized rehabilitation insights. By unifying signal processing, efficient deep learning, and interpretability, X-LiteHAR advances real-time HAR for clinical and wearable applications.
- 16 April 2025
- Published by Taylor & Francis in Remote Sensing Letters
- Vol. 16 (6) , 654-664
Mapping urban heat islands in Pune, India: ecological impacts and environmental challenges
- 16 April 2025
- Published by Taylor & Francis in Remote Sensing Letters
- Vol. 16 (6) , 654-664
Mapping urban heat islands in Pune, India: ecological impacts and environmental challenges