Suriya Prakash Muthukrishnan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Suriya Prakash Muthukrishnan has a total of 91 co-authors. Here are the top co-authors by number of shared publications:
- Ratna Sharma
Stress and Cognitive Electroimaging Laboratory, Department of Physiology, All India Institute of Medical Sciences, New Delhi, India
12 shared publications Sunaina Soni
Department of Physiology, All India Institute of Medical Sciences, New Delhi, India
9 shared publications- Simran Kaur
Stress and Cognitive Electroimaging Laboratory, Department of Physiology, All India Institute of Medical Sciences, New Delhi, India
8 shared publications - Lalan Kumar
Indian Institute of Technology Delhi
7 shared publications - Sitikantha Roy
Department of Applied Mechanics, Indian Institute of Technology Delhi, New Delhi, India
6 shared publications - Prashant Tayade
Department of Physiology, All India Institute of Medical Sciences, New Delhi, India
4 shared publications
Latest Publications
- Open Access
- 11 September 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1976480
Editorial: Complex network dynamics of cognitive processing in health and disease: current knowledge and future research
(Faskowitz et al., 2020;Novelli and Razi, 2022;Finc et al., 2026). Advances in neuroimaging and electrophysiology-including EEG, fMRI, MEG, PET, and multimodal combinationshave enabled researchers to characterize brain-network organization across complementary spatial and temporal scales, although each modality captures only part of the underlying dynamics (Lake and Higley, 2022). The evidence spans millisecond EEG states, moment-tomoment fMRI co-fluctuations, multimodal disease-network mapping, motor preparation, and neuroimmune/genomic interactions. The strongest convergence is that disease or increased cognitive demand changes network state transitions, not just regional activity. Simultaneously, advanced computational methods such as graph theory and machine learning offer new frameworks for understanding brain connectivity complexity (Mohammadi and Karwowski, 2024;Yang et al., 2025). Two contributions in this collection demonstrate that EEG microstates -brief, quasi-stable patterns of global scalp electrical activity lasting 80-120 milliseconds -offer a uniquely sensitive window onto large-scale network dynamics (Michel et al., 2024). Across these five contributions, several cross-cutting themes emerge. First, cognitive health depends on flexibility and on coordinated integration and segregation (Barabino et al., 2024).The DMN appears as a recurring locus of disruption -in MCI, SCD, PD with ICDs, and PTSD -underscoring its centrality to self-referential processing, memory, and cognitive control across the disease spectrum (Smallwood et al., 2021;Sanz-Morales and Melero, 2024).Interestingly, network dynamics emerge as predictive and context-sensitive features. Second, methodological innovation is driving discovery: edge-centric connectivity (Novelli and Razi, 2022), microstate analysis, and FCNM each offer complementary perspectives that conventional region-of-interest or seed-based approaches cannot provide. Third, the integration of genetics, immunology, and neuroimaging signals a necessary move toward multimodal and multiscale models of cognition and disease.Looking ahead, longitudinal designs are critically required to determine whether the network signatures identified herein -including altered microstate dynamics, disrupted co-fluctuation patterns, and subcortical-cortical imbalances -precede clinical symptom onset and predict disease trajectory. Larger and more diverse cohorts will be essential to validate these biomarkers across different populations. The translation of EEG microstate and connectivity biomarkers into clinical tools -for early diagnosis, patient stratification, and treatment monitoring -represents a particularly exciting frontier (Michel et al., 2024). Finally, the integration of non-invasive brain stimulation approaches (TMS, tDCS with real-time networkstate decoding offers a pathway toward genuinely closed-loop, network-targeted interventions (Wischnewski et al., 2024;Casazza et al., 2026).
- 11 September 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1976480
Open AccessEditorial: Complex network dynamics of cognitive processing in health and disease: current knowledge and future research
(Faskowitz et al., 2020;Novelli and Razi, 2022;Finc et al., 2026). Advances in neuroimaging and electrophysiology-including EEG, fMRI, MEG, PET, and multimodal combinationshave enabled researchers to characterize brain-network organization across complementary spatial and temporal scales, although each modality captures only part of the underlying dynamics (Lake and Higley, 2022). The evidence spans millisecond EEG states, moment-tomoment fMRI co-fluctuations, multimodal disease-network mapping, motor preparation, and neuroimmune/genomic interactions. The strongest convergence is that disease or increased cognitive demand changes network state transitions, not just regional activity. Simultaneously, advanced computational methods such as graph theory and machine learning offer new frameworks for understanding brain connectivity complexity (Mohammadi and Karwowski, 2024;Yang et al., 2025). Two contributions in this collection demonstrate that EEG microstates -brief, quasi-stable patterns of global scalp electrical activity lasting 80-120 milliseconds -offer a uniquely sensitive window onto large-scale network dynamics (Michel et al., 2024). Across these five contributions, several cross-cutting themes emerge. First, cognitive health depends on flexibility and on coordinated integration and segregation (Barabino et al., 2024).The DMN appears as a recurring locus of disruption -in MCI, SCD, PD with ICDs, and PTSD -underscoring its centrality to self-referential processing, memory, and cognitive control across the disease spectrum (Smallwood et al., 2021;Sanz-Morales and Melero, 2024).Interestingly, network dynamics emerge as predictive and context-sensitive features. Second, methodological innovation is driving discovery: edge-centric connectivity (Novelli and Razi, 2022), microstate analysis, and FCNM each offer complementary perspectives that conventional region-of-interest or seed-based approaches cannot provide. Third, the integration of genetics, immunology, and neuroimaging signals a necessary move toward multimodal and multiscale models of cognition and disease.Looking ahead, longitudinal designs are critically required to determine whether the network signatures identified herein -including altered microstate dynamics, disrupted co-fluctuation patterns, and subcortical-cortical imbalances -precede clinical symptom onset and predict disease trajectory. Larger and more diverse cohorts will be essential to validate these biomarkers across different populations. The translation of EEG microstate and connectivity biomarkers into clinical tools -for early diagnosis, patient stratification, and treatment monitoring -represents a particularly exciting frontier (Michel et al., 2024). Finally, the integration of non-invasive brain stimulation approaches (TMS, tDCS with real-time networkstate decoding offers a pathway toward genuinely closed-loop, network-targeted interventions (Wischnewski et al., 2024;Casazza et al., 2026).
- 3 September 2026
- Published by Elsevier in Seminars in Orthodontics
Orthodontic correction of a dentofacial deformity secondary to facial burn and neck contracture. A case report
- 3 September 2026
- Published by Elsevier in Seminars in Orthodontics
Orthodontic correction of a dentofacial deformity secondary to facial burn and neck contracture. A case report
- 4 June 2026
- Published by Springer Nature
- p. 290-301
EEG-Based 90-Degree Turn Intention Detection for Brain-Computer Interface
- 4 June 2026
- Published by Springer Nature
- p. 290-301
EEG-Based 90-Degree Turn Intention Detection for Brain-Computer Interface
- Open Access
- 23 March 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1784913
Pre-movement EEG microstates reflect intended lifted load of volitional movement
Introduction: Load estimation is one of the essential parameters for assistive robotic control in cases of rehabilitation. The high temporal resolution of the Electroencephalography (EEG) technique makes it the best tool to resolve the temporal dynamics of movement intention and planning. The quasi-stable scalp electrical potential topography represented by the EEG microstates could assess the real-time information processing in the brain for controlling assistive devices. We hypothesize that the EEG microstate preceding the movement could reflect the increasing load during a biceps curl movement. Methods: Ten healthy participants performed biceps curl movements, while their brain activity and muscle activation was recorded using EEG and EMG. Results: Eight microstate maps were found to represent the functional brain state before the movements. Two pre-movement microstate maps were found to reflect the load increments. The source maxima of these two reflective microstates maps were localized at the right insula and cingulate gyrus. Discussion: Our results imply that the load increments of volitional movement could be reflected by the pre-movement EEG microstates.
- 23 March 2026
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 20, 1784913
Open AccessPre-movement EEG microstates reflect intended lifted load of volitional movement
Introduction: Load estimation is one of the essential parameters for assistive robotic control in cases of rehabilitation. The high temporal resolution of the Electroencephalography (EEG) technique makes it the best tool to resolve the temporal dynamics of movement intention and planning. The quasi-stable scalp electrical potential topography represented by the EEG microstates could assess the real-time information processing in the brain for controlling assistive devices. We hypothesize that the EEG microstate preceding the movement could reflect the increasing load during a biceps curl movement. Methods: Ten healthy participants performed biceps curl movements, while their brain activity and muscle activation was recorded using EEG and EMG. Results: Eight microstate maps were found to represent the functional brain state before the movements. Two pre-movement microstate maps were found to reflect the load increments. The source maxima of these two reflective microstates maps were localized at the right insula and cingulate gyrus. Discussion: Our results imply that the load increments of volitional movement could be reflected by the pre-movement EEG microstates.
- Open Access
Load-Aware Calibration of EMG-Driven Musculoskeletal Models for Accurate and Generalizable Joint Torque Estimation
Accurate EMG-driven musculoskeletal (MSK) modeling is critical for biomechanics, rehabilitation, and assistive technology. However, most models calibrate parameters under a single load, ignoring the fact that tasks with similar kinematics may differ in mechanical demand. This study introduces a load-aware calibration framework to improve joint torque prediction accuracy and generalizability. Surface EMG and joint kinematics were recorded from eleven participants during elbow flexion-extension under 0, 2, and 4\,kg loads. We evaluated three calibration strategies (load-specific, global, cross-load) and three optimization frameworks (simulated annealing (SA), particle swarm optimization (PSO), and hybrid PSO-pattern search (PSO-PS)). Results indicate that load-specific calibration significantly improves performance, with lower RMSE and higher correlation ($r > 0.75$). Parameters related to muscle force, fiber length, and activation dynamics showed high load sensitivity. PSO-based methods yielded more consistent and physiologically plausible estimates than simulated annealing. The proposed framework enables MSK models to distinguish between visually similar but mechanically distinct tasks, supporting robust subject-specific modeling for clinical and real-world applications.
Open AccessLoad-Aware Calibration of EMG-Driven Musculoskeletal Models for Accurate and Generalizable Joint Torque Estimation
Accurate EMG-driven musculoskeletal (MSK) modeling is critical for biomechanics, rehabilitation, and assistive technology. However, most models calibrate parameters under a single load, ignoring the fact that tasks with similar kinematics may differ in mechanical demand. This study introduces a load-aware calibration framework to improve joint torque prediction accuracy and generalizability. Surface EMG and joint kinematics were recorded from eleven participants during elbow flexion-extension under 0, 2, and 4\,kg loads. We evaluated three calibration strategies (load-specific, global, cross-load) and three optimization frameworks (simulated annealing (SA), particle swarm optimization (PSO), and hybrid PSO-pattern search (PSO-PS)). Results indicate that load-specific calibration significantly improves performance, with lower RMSE and higher correlation ($r > 0.75$). Parameters related to muscle force, fiber length, and activation dynamics showed high load sensitivity. PSO-based methods yielded more consistent and physiologically plausible estimates than simulated annealing. The proposed framework enables MSK models to distinguish between visually similar but mechanically distinct tasks, supporting robust subject-specific modeling for clinical and real-world applications.
- 15 July 2025
- Published by Elsevier in Neurocomputing
- Vol. 651
PiMAN: A Physics-informed Motion Prediction Network using sEMG signal features for human movement parameters
- 15 July 2025
- Published by Elsevier in Neurocomputing
- Vol. 651
PiMAN: A Physics-informed Motion Prediction Network using sEMG signal features for human movement parameters
- Open Access
- 13 March 2025
- Published by Frontiers Media SA in Frontiers in Human Neuroscience
- Vol. 19, 1582214
Editorial: Neural mechanisms of motor planning in assisted voluntary movement
Improving humanoid robot hand dexterity enhances their ability to perform precise tasks, such as surgical assistance and aiding individuals with disabilities. The brain simplifies voluntary movement by organizing muscle and joint activations into coordinated patterns called "synergies," reducing the complexity of controlling the hand's numerous degrees of freedom.Studying these synergies provides insights into brain-hand communication, motor disorders, and robotic control. Researchers have explored kinematic synergies (Grinyagin et al., 2005;Freitas et al., 2006), to better understand the neural activity that mediates musculoskeletal mechanics and behavioral goals. Olikkal et al. extracted kinematic synergies from 33 American Sign Language hand gestures using an RGB camera, MediaPipe, Gaussian functions, and Principal component analysis (PCA), achieving 95.7% accuracy. This synergy-based approach simplifies motion retargeting, offering potential for assistive robotics.Accident analyses highlight run-off road incidents as a major cause of fatalities, yet drivers' brain responses during such events remain poorly understood. Brain recordings capture overlapping processes such as motor control, visual processing, and error monitoring, making their individual contributions unclear. While studies have separated visual and motor components (Walter et al., 2001;Horikawa et al., 2005;Garcia et al., 2017), disentangling these processes remains challenging. Pulferer et al. addressed this by employing passive and active steering in error-free and error-prone conditions, demonstrating that distinct subprocesses can be separated using time-locked analyses of EEG data. Findings revealed increased frontocentral activity and information flow during execution, linked to performance monitoring in the caudal anterior cingulate cortex.Reaching movements, fundamental to daily life and rehabilitation, follow Fitts' Law, which predicts longer movement times for more difficult tasks (Fitts, 1954). While Fitts' Law has enhanced human-machine interactions in neurorehabilitative devices (Zimmerli et al., 2012), the association between cortical activation and task demands remains unclear, hindering optimal therapeutic parameter selection (e.g., dosage, repetition, difficulty). Ji et al. examined whether motor cortex activity correlates with index-of-difficulty in speed-accuracy reaching tasks. Healthy subjects performed 2D reaching movements with ID levels using a rehabilitation robot while fNIRS recorded cortical responses. While kinematic data aligned with Fitts' Law, motor cortex activity showed no direct correlation with task difficulty, implying the influence of additional factors such as muscle activation.Physical exercise enhances brain plasticity, crucial for functional reorganization of the lesioned cortex and motor recovery in patients with motor impairment (Kokotilo et al., 2009;James and McGlinchey, 2022). However, how different movement patterns influence somatosensory cortex reorganization across various stages of neurorehabilitation remains poorly understood.Understanding their specific effects on sensory-motor cortex activity can refine training dosages (type, time, intensity) and deepen our understanding of exercise mechanisms. To explore the impact of exercise training modes on sensory and motor-related cortex excitability, Li et al. used fNIRS to study cortical activity in healthy participants during passive, active, and resistance tasks with an upper-limb robotic device. Active movement showed higher contralateral M1 activation, while resistance exercise activated both hemispheres more extensively. While these findings provide valuable insights, further research is needed to refine exercise therapy strategies using assistive technologies.Human movement is defined by kinematic and kinetic attributes. Kinematics, or "high-level control," governs motion parameters like location, direction, velocity, and acceleration, shaping the desired trajectory. Kinetics, or "low-level control," related to the control of individual muscles and forces. Multiple trajectories can achieve the same goal-oriented movement, and research has examined how the sensorimotor cortex represents these features (Branco et al., 2019;Zhou et al., 2022). However, how the brain optimally executes voluntary movements remains a major challenge. Ghosh et al. reviewed the neural correlates predictive of upper limb motor intention and kinematics. This review also highlights the potential of closed-loop EEGbased BMIs to promote long-term rehabilitation, neural plasticity, and motor recovery.In conclusion, the articles in this Research Topic provide insights into the neural mechanisms of motor planning in assisted voluntary movement and advancements in BMI-controlled assistive robotics. However, challenges remain, including the need for larger studies, standardized methodologies, and rigorous bias assessments. Future BMI systems for motor rehabilitation should focus on integrating multiple physiological signals, ensuring long-term stability, improving user engagement, and enhancing sensory feedback. Continued research and clinical trials are essential to developing effective BMI systems and improving the quality of life for patients with motor impairments.
- 13 March 2025
- Published by Frontiers Media SA in Frontiers in Human Neuroscience
- Vol. 19, 1582214
Open AccessEditorial: Neural mechanisms of motor planning in assisted voluntary movement
Improving humanoid robot hand dexterity enhances their ability to perform precise tasks, such as surgical assistance and aiding individuals with disabilities. The brain simplifies voluntary movement by organizing muscle and joint activations into coordinated patterns called "synergies," reducing the complexity of controlling the hand's numerous degrees of freedom.Studying these synergies provides insights into brain-hand communication, motor disorders, and robotic control. Researchers have explored kinematic synergies (Grinyagin et al., 2005;Freitas et al., 2006), to better understand the neural activity that mediates musculoskeletal mechanics and behavioral goals. Olikkal et al. extracted kinematic synergies from 33 American Sign Language hand gestures using an RGB camera, MediaPipe, Gaussian functions, and Principal component analysis (PCA), achieving 95.7% accuracy. This synergy-based approach simplifies motion retargeting, offering potential for assistive robotics.Accident analyses highlight run-off road incidents as a major cause of fatalities, yet drivers' brain responses during such events remain poorly understood. Brain recordings capture overlapping processes such as motor control, visual processing, and error monitoring, making their individual contributions unclear. While studies have separated visual and motor components (Walter et al., 2001;Horikawa et al., 2005;Garcia et al., 2017), disentangling these processes remains challenging. Pulferer et al. addressed this by employing passive and active steering in error-free and error-prone conditions, demonstrating that distinct subprocesses can be separated using time-locked analyses of EEG data. Findings revealed increased frontocentral activity and information flow during execution, linked to performance monitoring in the caudal anterior cingulate cortex.Reaching movements, fundamental to daily life and rehabilitation, follow Fitts' Law, which predicts longer movement times for more difficult tasks (Fitts, 1954). While Fitts' Law has enhanced human-machine interactions in neurorehabilitative devices (Zimmerli et al., 2012), the association between cortical activation and task demands remains unclear, hindering optimal therapeutic parameter selection (e.g., dosage, repetition, difficulty). Ji et al. examined whether motor cortex activity correlates with index-of-difficulty in speed-accuracy reaching tasks. Healthy subjects performed 2D reaching movements with ID levels using a rehabilitation robot while fNIRS recorded cortical responses. While kinematic data aligned with Fitts' Law, motor cortex activity showed no direct correlation with task difficulty, implying the influence of additional factors such as muscle activation.Physical exercise enhances brain plasticity, crucial for functional reorganization of the lesioned cortex and motor recovery in patients with motor impairment (Kokotilo et al., 2009;James and McGlinchey, 2022). However, how different movement patterns influence somatosensory cortex reorganization across various stages of neurorehabilitation remains poorly understood.Understanding their specific effects on sensory-motor cortex activity can refine training dosages (type, time, intensity) and deepen our understanding of exercise mechanisms. To explore the impact of exercise training modes on sensory and motor-related cortex excitability, Li et al. used fNIRS to study cortical activity in healthy participants during passive, active, and resistance tasks with an upper-limb robotic device. Active movement showed higher contralateral M1 activation, while resistance exercise activated both hemispheres more extensively. While these findings provide valuable insights, further research is needed to refine exercise therapy strategies using assistive technologies.Human movement is defined by kinematic and kinetic attributes. Kinematics, or "high-level control," governs motion parameters like location, direction, velocity, and acceleration, shaping the desired trajectory. Kinetics, or "low-level control," related to the control of individual muscles and forces. Multiple trajectories can achieve the same goal-oriented movement, and research has examined how the sensorimotor cortex represents these features (Branco et al., 2019;Zhou et al., 2022). However, how the brain optimally executes voluntary movements remains a major challenge. Ghosh et al. reviewed the neural correlates predictive of upper limb motor intention and kinematics. This review also highlights the potential of closed-loop EEGbased BMIs to promote long-term rehabilitation, neural plasticity, and motor recovery.In conclusion, the articles in this Research Topic provide insights into the neural mechanisms of motor planning in assisted voluntary movement and advancements in BMI-controlled assistive robotics. However, challenges remain, including the need for larger studies, standardized methodologies, and rigorous bias assessments. Future BMI systems for motor rehabilitation should focus on integrating multiple physiological signals, ensuring long-term stability, improving user engagement, and enhancing sensory feedback. Continued research and clinical trials are essential to developing effective BMI systems and improving the quality of life for patients with motor impairments.
- Open Access
- 6 February 2025
- Published by Frontiers Media SA in Frontiers in Human Neuroscience
- Vol. 19, 1532783
Decoding the brain-machine interaction for upper limb assistive technologies: advances and challenges
Understanding how the brain encodes upper limb movements is crucial for developing control mechanisms in assistive technologies. Advances in assistive technologies, particularly Brain-machine Interfaces (BMIs), highlight the importance of decoding motor intentions and kinematics for effective control. EEG-based BMI systems show promise due to their non-invasive nature and potential for inducing neural plasticity, enhancing motor rehabilitation outcomes. While EEG-based BMIs show potential for decoding motor intention and kinematics, studies indicate inconsistent correlations with actual or planned movements, posing challenges for achieving precise and reliable prosthesis control. Further, the variability in predictive EEG patterns across individuals necessitates personalized tuning to improve BMI efficiency. Integrating multiple physiological signals could enhance BMI precision and reliability, paving the way for more effective motor rehabilitation strategies. Studies have shown that brain activity adapts to gravitational and inertial constraints during movement, highlighting the critical role of neural adaptation to biomechanical changes in creating control systems for assistive devices. This review aims to provide a comprehensive overview of recent progress in deciphering neural activity patterns associated with both physiological and assisted upper limb movements, highlighting avenues for future exploration in neurorehabilitation and brain-machine interface development.
- 6 February 2025
- Published by Frontiers Media SA in Frontiers in Human Neuroscience
- Vol. 19, 1532783
Open AccessDecoding the brain-machine interaction for upper limb assistive technologies: advances and challenges
Understanding how the brain encodes upper limb movements is crucial for developing control mechanisms in assistive technologies. Advances in assistive technologies, particularly Brain-machine Interfaces (BMIs), highlight the importance of decoding motor intentions and kinematics for effective control. EEG-based BMI systems show promise due to their non-invasive nature and potential for inducing neural plasticity, enhancing motor rehabilitation outcomes. While EEG-based BMIs show potential for decoding motor intention and kinematics, studies indicate inconsistent correlations with actual or planned movements, posing challenges for achieving precise and reliable prosthesis control. Further, the variability in predictive EEG patterns across individuals necessitates personalized tuning to improve BMI efficiency. Integrating multiple physiological signals could enhance BMI precision and reliability, paving the way for more effective motor rehabilitation strategies. Studies have shown that brain activity adapts to gravitational and inertial constraints during movement, highlighting the critical role of neural adaptation to biomechanical changes in creating control systems for assistive devices. This review aims to provide a comprehensive overview of recent progress in deciphering neural activity patterns associated with both physiological and assisted upper limb movements, highlighting avenues for future exploration in neurorehabilitation and brain-machine interface development.
- 25 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73  (00189456) , 1-13
NeuroAiR: Deep Learning Framework for Airwriting Recognition From Scalp-Recorded Neural Signals
Airwriting recognition is a task that involves identifying letters written in free space using finger movement. It is a special case of gesture recognition, where gestures correspond to letters in a specific language. Electroencephalography (EEG) is a non-invasive technique for recording brain activity and has been widely used in brain-computer interface applications. Leveraging EEG signals for airwriting recognition offers a promising alternative input method for Human-Computer Interaction. One key advantage of airwriting recognition is that users do not need to learn new gestures. By concatenating recognized letters, a wide range of words can be formed, making it applicable to a broader population. However, there has been limited research in the recognition of airwriting using EEG signals, which forms the core focus of this study. The NeuroAiR dataset comprising EEG signals recorded during writing English uppercase alphabets is first constructed. Various features are then explored in conjunction with different deep learning models to achieve accurate airwriting recognition. These features include processed EEG data, Independent Component Analysis components, source-domain-based scout time series, and spherical and head harmonic decomposition-based features. Furthermore, the impact of different EEG frequency bands on system performance is comprehensively investigated. The highest accuracy achieved in this study is 44.04% using Independent Component Analysis components and the EEGNet classification model. The results highlight the potential of EEG-based airwriting recognition as a user-friendly modality for alternative input methods in Human-Computer Interaction applications. This research sets a strong baseline for future advancements and demonstrates the viability and utility of EEG-based airwriting recognition.
- 25 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73  (00189456) , 1-13
NeuroAiR: Deep Learning Framework for Airwriting Recognition From Scalp-Recorded Neural Signals
Airwriting recognition is a task that involves identifying letters written in free space using finger movement. It is a special case of gesture recognition, where gestures correspond to letters in a specific language. Electroencephalography (EEG) is a non-invasive technique for recording brain activity and has been widely used in brain-computer interface applications. Leveraging EEG signals for airwriting recognition offers a promising alternative input method for Human-Computer Interaction. One key advantage of airwriting recognition is that users do not need to learn new gestures. By concatenating recognized letters, a wide range of words can be formed, making it applicable to a broader population. However, there has been limited research in the recognition of airwriting using EEG signals, which forms the core focus of this study. The NeuroAiR dataset comprising EEG signals recorded during writing English uppercase alphabets is first constructed. Various features are then explored in conjunction with different deep learning models to achieve accurate airwriting recognition. These features include processed EEG data, Independent Component Analysis components, source-domain-based scout time series, and spherical and head harmonic decomposition-based features. Furthermore, the impact of different EEG frequency bands on system performance is comprehensively investigated. The highest accuracy achieved in this study is 44.04% using Independent Component Analysis components and the EEGNet classification model. The results highlight the potential of EEG-based airwriting recognition as a user-friendly modality for alternative input methods in Human-Computer Interaction applications. This research sets a strong baseline for future advancements and demonstrates the viability and utility of EEG-based airwriting recognition.
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73  (00189456) , 1-11
BiCurNet: Premovement EEG-Based Neural Decoder for Biceps Curl Trajectory Estimation
Kinematic parameter (KP) estimation from early electroencephalogram (EEG) signals is essential for positive augmentation using wearable robots. However, surface EEG-based early KP estimation studies are sparse in the literature. In this study, simultaneous surface EEG and kinematics data of five participants is collected during the biceps-curl motor task. The feasibility of early estimation of KPs is demonstrated using brain source imaging (BSI). Discrete wavelet transform (DWT) is utilized for sub-band extraction from pre-processed EEG signals. Further, spherical and head harmonics domain features are extracted from sub-bands of the EEG signals. A deep-learning based decoding model, BiCurNet, is proposed for early KP estimation using spatial and harmonics domain EEG features during the biceps-curl task. The proposed model utilizes lightweight architecture with depth-wise separable convolution layers and a customized attention module. The best Pearson correlation coefficient (PCC) between the estimated and actual trajectory of 0.7 is achieved when combined EEG features (spatial and harmonics domain) in the delta band are utilized. Intra- and Inter-subject performance analyses are performed to evaluate the subject-adaptability of the proposed decoding model. The performance of the proposed BiCurNet is compared with the existing multi-linear regression (mLR) counterpart. The robustness of the proposed model is additionally illustrated using an ablation study. The robust performance and lightweight architecture will facilitate the real-time implementation of the model with deployment on a microcontroller to control BCI-based wearable robots.
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73  (00189456) , 1-11
BiCurNet: Premovement EEG-Based Neural Decoder for Biceps Curl Trajectory Estimation
Kinematic parameter (KP) estimation from early electroencephalogram (EEG) signals is essential for positive augmentation using wearable robots. However, surface EEG-based early KP estimation studies are sparse in the literature. In this study, simultaneous surface EEG and kinematics data of five participants is collected during the biceps-curl motor task. The feasibility of early estimation of KPs is demonstrated using brain source imaging (BSI). Discrete wavelet transform (DWT) is utilized for sub-band extraction from pre-processed EEG signals. Further, spherical and head harmonics domain features are extracted from sub-bands of the EEG signals. A deep-learning based decoding model, BiCurNet, is proposed for early KP estimation using spatial and harmonics domain EEG features during the biceps-curl task. The proposed model utilizes lightweight architecture with depth-wise separable convolution layers and a customized attention module. The best Pearson correlation coefficient (PCC) between the estimated and actual trajectory of 0.7 is achieved when combined EEG features (spatial and harmonics domain) in the delta band are utilized. Intra- and Inter-subject performance analyses are performed to evaluate the subject-adaptability of the proposed decoding model. The performance of the proposed BiCurNet is compared with the existing multi-linear regression (mLR) counterpart. The robustness of the proposed model is additionally illustrated using an ablation study. The robust performance and lightweight architecture will facilitate the real-time implementation of the model with deployment on a microcontroller to control BCI-based wearable robots.