Naveen Gehlot
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Naveen Gehlot has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Rajesh Kumar
Department of Electrical Engineering, Malaviya National Institute of Technology, Jaipur 302017, Rajasthan, India
6 shared publications - Ankit Vijayvargiya
School of Technology, Dhirubhai Ambani University
3 shared publications - Ashutosh Jena
Department of Electrical Engineering, Malaviya National Institute of Technology Jaipur, Jaipur, India
2 shared publications - Mahipal Bukya
Department of Electrical and Electronics Engineering, Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education, Manipal, India
2 shared publications Rajesh Kumar
2 shared publications- Akhil Ranjan Garg
J.N.V University
1 shared publication
Latest Publications
- Open Access
- 26 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 32146-32157
BAYES-GAIT: A Bayesian Deep Neural Network for Continuous Gait Reference Trajectory Generation Using an Enhanced Loss Function
Background: Adaptive gait trajectory prediction is essential to achieve natural and stable locomotion in prosthetic limbs and legged robots, particularly under varied conditions such as changing inclines and walking speeds. Although deep neural networks have shown potential for learning complex joint gait trajectory patterns, conventional deep neural networks often suffer from prediction bias and lack mechanisms for uncertainty estimation, which limits their reliability in dynamic or unseen environments. Methods: This study proposes a Bayesian deep neural network framework for predicting continuous joint trajectories in the sagittal plane across diverse ambulation tasks. The approach integrates Bayesian inference to estimate predictive uncertainty and enhance model generalization. A new hybrid loss function is introduced that combines the standard error of the dataset with variational mean absolute error or mean squared error, allowing the model to leverage low-variance gait segments more effectively during training. The framework is trained and evaluated using a dataset that includes varying walking speeds and inclines. Results: The proposed Bayesian model outperforms traditional state-of-the-art methods, particularly deep neural networks and finite state machine baselines, and shows up to 15% and 25% improvements, respectively, in joint trajectory prediction accuracy. Moreover, the Bayesian deep neural network provides 95% prediction intervals, indicating strong reliability and robustness under uncertain conditions. Significance: By incorporating uncertainty quantification and dataset-driven variance into the learning process, this study offers a more reliable and adaptive approach to gait modeling. These improvements support enhanced gait control in assistive technologies and legged robotic systems operating in variable terrain conditions in the real world.
- 26 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 32146-32157
Open AccessBAYES-GAIT: A Bayesian Deep Neural Network for Continuous Gait Reference Trajectory Generation Using an Enhanced Loss Function
Background: Adaptive gait trajectory prediction is essential to achieve natural and stable locomotion in prosthetic limbs and legged robots, particularly under varied conditions such as changing inclines and walking speeds. Although deep neural networks have shown potential for learning complex joint gait trajectory patterns, conventional deep neural networks often suffer from prediction bias and lack mechanisms for uncertainty estimation, which limits their reliability in dynamic or unseen environments. Methods: This study proposes a Bayesian deep neural network framework for predicting continuous joint trajectories in the sagittal plane across diverse ambulation tasks. The approach integrates Bayesian inference to estimate predictive uncertainty and enhance model generalization. A new hybrid loss function is introduced that combines the standard error of the dataset with variational mean absolute error or mean squared error, allowing the model to leverage low-variance gait segments more effectively during training. The framework is trained and evaluated using a dataset that includes varying walking speeds and inclines. Results: The proposed Bayesian model outperforms traditional state-of-the-art methods, particularly deep neural networks and finite state machine baselines, and shows up to 15% and 25% improvements, respectively, in joint trajectory prediction accuracy. Moreover, the Bayesian deep neural network provides 95% prediction intervals, indicating strong reliability and robustness under uncertain conditions. Significance: By incorporating uncertainty quantification and dataset-driven variance into the learning process, this study offers a more reliable and adaptive approach to gait modeling. These improvements support enhanced gait control in assistive technologies and legged robotic systems operating in variable terrain conditions in the real world.
- 19 January 2026
- Published by Taylor & Francis
- p. 166-187
1D self-organizing feature map technique for diagnosis of COVID-19 infection
- 19 January 2026
- Published by Taylor & Francis
- p. 166-187
1D self-organizing feature map technique for diagnosis of COVID-19 infection
- 18 January 2026
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 167
Spiking neural network-based energy-efficient framework for real-time robotic arm manipulation
- 18 January 2026
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 167
Spiking neural network-based energy-efficient framework for real-time robotic arm manipulation
- 31 August 2025
- Published by Springer Nature in SN Computer Science
- Vol. 6 (7) , 1-19
1D Convolutional Neural Architecture Search for sEMG Hand Gesture Recognition
- 31 August 2025
- Published by Springer Nature in SN Computer Science
- Vol. 6 (7) , 1-19
1D Convolutional Neural Architecture Search for sEMG Hand Gesture Recognition
- 20 May 2025
- Published by Springer Nature
- p. 129-140
DWMS-HGR: 1D CNN for Hand Gesture Recognition on Normalized sEMG Signals with Sliding Window
- 20 May 2025
- Published by Springer Nature
- p. 129-140
DWMS-HGR: 1D CNN for Hand Gesture Recognition on Normalized sEMG Signals with Sliding Window
- Open Access
- 3 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 145540-145549
Generating Topic-Agnostic Conversations With LLMs
Conversational systems are important applications of Artificial Intelligence, encompassing a wide variety of implementations, from rule-based systems to complex systems using Natural Language Processing, Deep Neural Networks, and Transformer Architectures. With the growth of these implementations, the quality of conversational data has become a concern. Many attempts have been made to generate such data, focusing primarily on topical conversations. This article presents a generalized framework moving from generating topical conversation towards topic-agnostic conversational data consisting of three Large Language Model instances. Two of these models interact with each other to generate the conversation, while the third one plays the role of a judge to keep the conversation going. The synthetic data created by the proposed method exhibits higher quality and lower toxicity than four of the existing datasets (AmazonQA, Daily Dialog, Open Subtitles, and HUMOD) in terms of six performance measures, namely Toxicity, Severe Toxicity, Obscene, Threat, Insult, and Identity Attack. Compared to other datasets, the performance analysis of the generated data shows the lowest measures in terms of mean and maximum values. Specifically, the percentages for Toxicity, Obscene, Threat, Insult, and Identity Attack are 0.27%, 0.04%, 0.02%, 0.05%, and 0.02%, respectively, while the corresponding maximum values are 90.13%, 29.65%, 3.42%, 67.16% and 0.69%. The generated dataset also shows the maximum concentration, with 99.74% of the data in the range of 0-10% toxicity with just a few outliers.
- 3 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 145540-145549
Open AccessGenerating Topic-Agnostic Conversations With LLMs
Conversational systems are important applications of Artificial Intelligence, encompassing a wide variety of implementations, from rule-based systems to complex systems using Natural Language Processing, Deep Neural Networks, and Transformer Architectures. With the growth of these implementations, the quality of conversational data has become a concern. Many attempts have been made to generate such data, focusing primarily on topical conversations. This article presents a generalized framework moving from generating topical conversation towards topic-agnostic conversational data consisting of three Large Language Model instances. Two of these models interact with each other to generate the conversation, while the third one plays the role of a judge to keep the conversation going. The synthetic data created by the proposed method exhibits higher quality and lower toxicity than four of the existing datasets (AmazonQA, Daily Dialog, Open Subtitles, and HUMOD) in terms of six performance measures, namely Toxicity, Severe Toxicity, Obscene, Threat, Insult, and Identity Attack. Compared to other datasets, the performance analysis of the generated data shows the lowest measures in terms of mean and maximum values. Specifically, the percentages for Toxicity, Obscene, Threat, Insult, and Identity Attack are 0.27%, 0.04%, 0.02%, 0.05%, and 0.02%, respectively, while the corresponding maximum values are 90.13%, 29.65%, 3.42%, 67.16% and 0.69%. The generated dataset also shows the maximum concentration, with 99.74% of the data in the range of 0-10% toxicity with just a few outliers.
- Open Access
- 27 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 74185-74196
supDQN: Supervised Rewarding Strategy Driven Deep Q-Network for sEMG Signal Decontamination
The presence of muscles throughout the active parts of the body, such as the upper and lower limbs, makes electromyography-based human-machine interaction prevalent. However, muscle signals are stochastic and noisy, with noises being both regular and irregular. Irregular noises due to movements or electrical switching require dynamic filtering. Conventionally, filters are stacked, which unnecessarily trims and delays the signal. This study introduces a decontamination technique involving a supervised rewarding strategy to drive a deep Q-network-based agent (supDQN). It applies one of three filters to decontaminate a 1 sec long surface electromyography signal, which is dynamically contaminated. A machine learning agent identifies whether the signal after filtering is clean or noisy, generating a reward accordingly. The identification accuracy is enhanced by using a local interpretable model-agnostic explanation. The deep Q-network is guided by this reward to select the filter optimally while decontaminating a signal. The proposed filtering strategy is tested on four noise levels (−5 dB, −1 dB, +1 dB, +5 dB). supDQN filters the signal desirably when the signal-to-noise ratio (SNR) is between -5 dB to +1 dB but filters less desirably at high SNR (+5 dB). A normalized root mean square ($\Omega $ ) is formulated to depict the difference of the filtered signal from the ground truth. This is used to compare supDQN and conventional methods, including wavelet denoising with debauchies and symlet wavelet, high-order low-pass filter, notch filter, and high-pass filter. The proposed filtering strategy gives an average$\Omega $ value of 1.1974, which is lower than that of the conventional filters.- 27 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 74185-74196
Open AccesssupDQN: Supervised Rewarding Strategy Driven Deep Q-Network for sEMG Signal Decontamination
The presence of muscles throughout the active parts of the body, such as the upper and lower limbs, makes electromyography-based human-machine interaction prevalent. However, muscle signals are stochastic and noisy, with noises being both regular and irregular. Irregular noises due to movements or electrical switching require dynamic filtering. Conventionally, filters are stacked, which unnecessarily trims and delays the signal. This study introduces a decontamination technique involving a supervised rewarding strategy to drive a deep Q-network-based agent (supDQN). It applies one of three filters to decontaminate a 1 sec long surface electromyography signal, which is dynamically contaminated. A machine learning agent identifies whether the signal after filtering is clean or noisy, generating a reward accordingly. The identification accuracy is enhanced by using a local interpretable model-agnostic explanation. The deep Q-network is guided by this reward to select the filter optimally while decontaminating a signal. The proposed filtering strategy is tested on four noise levels (−5 dB, −1 dB, +1 dB, +5 dB). supDQN filters the signal desirably when the signal-to-noise ratio (SNR) is between -5 dB to +1 dB but filters less desirably at high SNR (+5 dB). A normalized root mean square ($\Omega $ ) is formulated to depict the difference of the filtered signal from the ground truth. This is used to compare supDQN and conventional methods, including wavelet denoising with debauchies and symlet wavelet, high-order low-pass filter, notch filter, and high-pass filter. The proposed filtering strategy gives an average$\Omega $ value of 1.1974, which is lower than that of the conventional filters.