Publications

Full text from these publications are available on HAL and via my google scholar profile.

2024

1. A systematic evaluation of Euclidean alignment with deep learning for EEG decoding Junqueira, Bruna, Aristimunha, Bruno, Chevallier, Sylvain, and De Camargo, Raphael Y Journal of Neural Engineering, 2024 [Abs] [HTML] [DOI]

   Abstract Objective: Electroencephalography signals are frequently used for various Brain–Computer interface (BCI) tasks. While deep learning (DL) techniques have shown promising results, they are hindered by the substantial data requirements. By leveraging data from multiple subjects, transfer learning enables more effective training of DL models. A technique that is gaining popularity is Euclidean alignment (EA) due to its ease of use, low computational complexity, and compatibility with DL models. However, few studies evaluate its impact on the training performance of shared and individual DL models. In this work, we systematically evaluate the effect of EA combined with DL for decoding BCI signals. Approach: We used EA as a pre-processing step to train shared DL models with data from multiple subjects and evaluated their transferability to new subjects. Main results: Our experimental results show that it improves decoding in the target subject by 4.33% and decreases convergence time by more than 70%. We also trained individual models for each subject to use as a majority-voting ensemble classifier. In this scenario, using EA improved the 3-model ensemble accuracy by 3.71%. However, when compared to the shared model with EA, the ensemble accuracy was 3.62% lower. Significance: EA succeeds in the task of improving transfer learning performance with DL models and, could be used as a standard pre-processing technique.

2. Growing Tiny Networks: Spotting Expressivity Bottlenecks and Fixing Them Optimally Verbockhaven, Manon, Chevallier, Sylvain, and Charpiat, Guillaume In arXiv, 2024 [HTML]

2023

1. Use of cognitive load measurements to design a new architecture of intelligent learning systems Zammouri, Amin, Moussa, Abdelaziz Ait, and Chevallier, Sylvain Expert Systems with Applications, 2023 [HTML] [DOI]
2. Accounting for endogenous effects in decision-making with a non-linear diffusion decision model Hoxha, Isabelle, Chevallier, Sylvain, Ciarchi, Matteo, Glasauer, Stefan, Delorme, Arnaud, and Amorim, Michel-Ange Scientific Reports, 2023 [Abs] [HTML] [DOI]

   Abstract The Drift-Diffusion Model (DDM) is widely accepted for two-alternative forced-choice decision paradigms thanks to its simple formalism and close fit to behavioral and neurophysiological data. However, this formalism presents strong limitations in capturing inter-trial dynamics at the single-trial level and endogenous influences. We propose a novel model, the non-linear Drift-Diffusion Model (nl-DDM), that addresses these issues by allowing the existence of several trajectories to the decision boundary. We show that the non-linear model performs better than the drift-diffusion model for an equivalent complexity. To give better intuition on the meaning of nl-DDM parameters, we compare the DDM and the nl-DDM through correlation analysis. This paper provides evidence of the functioning of our model as an extension of the DDM. Moreover, we show that the nl-DDM captures time effects better than the DDM. Our model paves the way toward more accurately analyzing across-trial variability for perceptual decisions and accounts for peri-stimulus influences.

3. Ensemble of Riemannian Classifiers for Multimodal Data: FUCONE Approach for M/EEG Data Corsi, Marie-Constance, Chevallier, Sylvain, Vico Fallani, Fabrizio, and Yger, Florian In ISBI 2023 - IEEE International Symposium on Biomedical Imaging, 2023 [HTML]
4. Holographic EEG: multi-view deep learning for BCI Carrara, Igor, Aristimunha, Bruno, Chevallier, Sylvain, Corsi, Marie-Constance, Papadopoulo, Théodore, and Corsi, Marie-Constance In CORTICO, 2023 [HTML]
5. Dimensionality Reduction and Frequency Bin Optimization To Improve a Riemannian-based Classification Pipeline Frateur, Rune, Chevallier, Sylvain, Yger, Florian, and Corsi, Marie-Constance In CORTICO, 2023 [HTML]
6. CONCERTO: Coherence & Functional Connectivity Graph Network Aristimunha, Bruno, Camargo, Raphael Yokoingawa De, Pinaya, Walter Hugo Lopez, Yger, Florian, Corsi, Marie-Constance, and Chevallier, Sylvain In CORTICO, 2023 [HTML]
7. Effectiveness of cross-frequency phase-amplitude covariance as additional features for Riemannian BCIs Yamamoto, Maria Sayu, Chevallier, Sylvain, and Lotte, Fabien In BCI Meeting, 2023 [HTML] [DOI]
8. Novel SPD matrix representations considering cross-frequency coupling for EEG classification using Riemannian geometry Yamamoto, Maria Sayu, Mellot, Apolline, Chevallier, Sylvain, and Lotte, Fabien In EUSIPCO 2023 - The 31st European Signal Processing Conference, 2023 [HTML]
9. Synthetic Sleep EEG Signal Generation using Latent Diffusion Models Aristimunha, Bruno, Camargo, Raphael Yokoingawa, Chevallier, Sylvain, Thomas, Adam G., Lucena, Oeslle, Cardoso, Jorge, Pinaya, Walter Hugo Lopez, and Dafflon, Jessica In DGM4H 2023 - 1st Workshop on Deep Generative Models for Health at NeurIPS 2023, 2023 [HTML]
10. Augmented SPDNet: Second-Order Neural Network for Motor Imagery-Based BCI Carrara, Igor, Aristimunha, Bruno, Corsi, Marie-Constance, Camargo, Raphael Yokoingawa De, Chevallier, Sylvain, and Papadopoulo, Théodore In Soph.IA Summit, 2023 [HTML]
11. Measuring Neuronal Avalanches to inform Brain-Computer Interfaces Corsi, Marie-Constance, Sorrentino, Pierpaolo, Schwartz, Denis P, George, Nathalie, Gollo, Leonardo L., Chevallier, Sylvain, Hugueville, Laurent, Kahn, Ari E, Dupont, Sophie, Bassett, Danielle S, Jirsa, Viktor, and Vico Fallani, Fabrizio iScience, 2023 [HTML] [DOI]
12. Self-administered non-invasive vagus nerve stimulation therapy for severe pharmacoresistant restless legs syndrome: outcomes at 6 months Hartley, Sarah, Bao, Guillaume, Russo, Ashley, Zagdoun, Marine, Chevallier, Sylvain, Lofaso, Frédéric, Leotard, Antoine, and Azabou, Eric Journal of Sleep Research, 2023 [HTML] [DOI]
13. EEG anticipatory activity depends on sensory modality Hoxha, Isabelle, Chevallier, Sylvain, Delorme, Arnaud, and Amorim, Michel-Ange bioRxiv, 2023 [Abs] [HTML] [DOI]

    Perceptual anticipation is known to have an impact on the reaction time of decisions. While anticipatory brain activity patterns have been identified in human EEG, in particular in the form of Contingent-Negative Variations (CNV), the single-trial neural signature of anticipation remains unexplored. Similarly, past studies have underlined an effect of pre-stimulus alpha-band activity on reaction times. Still, it remains unknown whether this activity is stimulus-specific or rather acts as a general indicator of readiness. This study aimed to decipher whether human participants expected a visual or an auditory stimulus at the single-trial level in both cued and uncued trials. We show that the CNV entails information about the expected upcoming stimulus, and the information content can be extracted at the single-trial level. Behavioral analyses additionally indicate the correct classification of uncued trials.Competing Interest StatementThe authors have declared no competing interest.

2022

1. 2021 BEETL Competition: Advancing Transfer Learning for Subject Independence and Heterogenous EEG Data Sets Wei, Xiaoxi, Faisal, A. Aldo, Grosse-Wentrup, Moritz, Gramfort, Alexandre, Chevallier, Sylvain, Jayaram, Vinay, Jeunet, Camille, Bakas, Stylianos, Ludwig, Siegfried, Barmpas, Konstantinos, Bahri, Mehdi, Panagakis, Yannis, Laskaris, Nikolaos, Adamos, Dimitrios A., Zafeiriou, Stefanos, Duong, William C., Gordon, Stephen M., Lawhern, Vernon J., Śliwowski, Maciej, Rouanne, Vincent, and Tempczyk, Piotr In Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track, 2022 [Abs] [HTML]

   Transfer learning and meta-learning offer some of the most promising avenues to unlock the scalability of healthcare and consumer technologies driven by biosignal data. This is because regular machine learning methods cannot generalise well across human subjects and handle learning from different, heterogeneously collected data sets, thus limiting the scale of training data available. On the other hand, the many developments in transfer- and meta-learning fields would benefit significantly from a real-world benchmark with immediate practical application. Therefore, we pick electroencephalography (EEG) as an exemplar for all the things that make biosignal data analysis a hard problem. We design two transfer learning challenges around a. clinical diagnostics and b. neurotechnology. These two challenges are designed to probe algorithmic performance with all the challenges of biosignal data, such as low signal-to-noise ratios, major variability among subjects, differences in the data recording sessions and techniques, and even between the specific BCI tasks recorded in the dataset. Task 1 is centred on the field of medical diagnostics, addressing automatic sleep stage annotation across subjects. Task 2 is centred on Brain-Computer Interfacing (BCI), addressing motor imagery decoding across both subjects and data sets. The successful 2021 BEETL competition with its over 30 competing teams and its 3 winning entries brought attention to the potential of deep transfer learning and combinations of set theory and conventional machine learning techniques to overcome the challenges. The results set a new state-of-the-art for the real-world BEETL benchmarks.

2. Functional Connectivity Ensemble Method to Enhance BCI Performance (FUCONE) Corsi, Marie-Constance, Chevallier, Sylvain, Fallani, Fabrizio De Vico, and Yger, Florian IEEE Transactions on Biomedical Engineering, 2022 [HTML] [DOI]
3. Riemannian geometry for combining functional connectivity metrics and covariance in BCI Chevallier, Sylvain, Corsi, Marie-Constance, Yger, Florian, and De Vico Fallani, Fabrizio Software Impacts, 2022 [HTML] [DOI]
4. Class-distinctiveness-based frequency band selection on the Riemannian manifold for oscillatory activity-based BCIs: preliminary results Yamamoto, Maria Sayu, Lotte, Fabien, Yger, Florian, and Chevallier, Sylvain In 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2022 [HTML] [DOI]
5. Noninvasive Vagus Nerve Stimulation: A New Therapeutic Approach for Pharmacoresistant Restless Legs Syndrome Hartley, Sarah, Bao, Guillaume, Zagdoun, Marine, Chevallier, Sylvain, Lofaso, Frédéric, Leotard, Antoine, and Azabou, Eric Neuromodulation: Technology at the Neural Interface, 2022 [HTML] [DOI]
6. Generating stimulus anticipation from stimulus and prediction history Hoxha, Isabelle, Bäker, Clark, Chevallier, Sylvain, Glasauer, Stefan, and Amorim, Michel-Ange In , 2022 [HTML] [DOI]
7. nl-DDM: a non-linear drift-diffusion model accounting for the dynamics of single-trial perceptual decisions Hoxha, Isabelle, Chevallier, Sylvain, Ciarchi, Matteo, Glasauer, Stefan, Delorme, Arnaud, and Amorim, Michel-Ange preprint, , 2022 [Abs] [HTML] [DOI]

   Abstract The Drift-Diffusion Model (DDM) is widely accepted for two-alternative forced-choice decision paradigms thanks to its simple formalism, straightforward interpretation, and close fit to behavioral and neurophysiological data. However, this formalism presents strong limitations to capture inter-trial dependency and dynamics at the single-trial level. We propose a novel model, the non-linear Drift-Diffusion Model (nl-DDM), that addresses these issues by allowing the existence of several trajectories to the decision boundary. We show that the fitting accuracy of our model is comparable to the accuracy of the DDM, with the non-linear model performing better than the drift-diffusion model for an equivalent complexity. To give better intuition on the meaning of nl-DDM parameters, we compare the DDM and the nl-DDM through correlation analysis. This paper provides evidence of the functioning of our model as an extension of the DDM. Our model paves the way toward more accurately analyzing single-trial dynamics for perceptual decisions and accounts for pre- and post-stimulus influences.

8. Un traitement innovant pour le syndrome des jambes sans repos pharmaco-résistant : la stimulation non invasive du nerf vague Hartley, Sarah, Bao, Guillaume, Zagdoun, Marine, Chevallier, Sylvain, Lajoie, Didier, Bohic, Mikaelle, Leotard, Antoine, and Azabou, Éric In Médecine du Sommeil, 2022 [HTML] [DOI]
9. User-specific frequency band and time segment selection with high class distinctiveness for Riemannian BCIs Yamamoto, Maria Sayu, Lotte, Fabien, Yger, Florian, and Chevallier, Sylvain In , 2022 [HTML]
10. End-to-end P300 BCI using Bayesian accumulation of Riemannian probabilities Barthélemy, Quentin, Chevallier, Sylvain, Bertrand-Lalo, Raphaëlle, and Clisson, Pierre Brain-Computer Interfaces, 2022 [HTML] [DOI]

2021

1. RIGOLETTO – RIemannian GeOmetry LEarning: applicaTion To cOnnectivity. A contribution to the Clinical BCI Challenge – WCCI2020 Corsi, Marie-Constance, Yger, Florian, Chevallier, Sylvain, and Noûs, Camille misc, , 2021 [Abs] [HTML] [DOI]

   This short technical report describes the approach submitted to the Clinical BCI Challenge-WCCI2020. This submission aims to classify motor imagery task from EEG signals and relies on Riemannian Geometry, with a twist. Instead of using the classical covariance matrices, we also rely on measures of functional connectivity. Our approach ranked 1st on the task 1 of the competition.

2. Minimizing Subject-dependent Calibration for BCI with Riemannian Transfer Learning Khazem, Salim, Chevallier, Sylvain, Barthélemy, Quentin, Haroun, Karim, and Noûs, Camille In , 2021 [Abs] [HTML] [DOI]

   Calibration is still an important issue for user experience in Brain-Computer Interfaces (BCI). Common experimental designs often involve a lengthy training period that raises the cognitive fatigue, before even starting to use the BCI. Reducing or suppressing this subject-dependent calibration is possible by relying on advanced machine learning techniques, such as transfer learning. Building on Riemannian BCI, we present a simple and effective scheme to train a classifier on data recorded from different subjects, to reduce the calibration while preserving good performances. The main novelty of this paper is to propose a unique approach that could be applied on very different paradigms. To demonstrate the robustness of this approach, we conducted a meta-analysis on multiple datasets for three BCI paradigms: motor imagery, eventrelated potentials (P300) and SSVEP. Relying on the MOABB open source framework to ensure the reproducibility of the experiments and the statistical analysis, the results clearly show that the proposed approach could be applied on any kind of BCI paradigm and in most of the cases to significantly improve the classifier reliability. We point out some key features to further improve transfer learning methods.

3. Subspace oddity - optimization on product of Stiefel manifolds for EEG data Yamamoto, Maria Sayu, Yger, Florian, and Chevallier, Sylvain In ICASSP, 2021 [Abs] [HTML] [DOI]

   Dimensionality reduction of high-dimensional electroencephalography (EEG) covariance matrices is crucial for effective utilization of Riemannian geometry in Brain-Computer Interfaces (BCI). In this paper, we propose a novel similaritybased classification method that relies on dimensionality reduction of EEG covariance matrices. Conventionally, the dimension of the original high-dimensional space is reduced by projecting into one low-dimensional space, and the similarity is learned only based on the single space. In contrast, our method, MUltiple SUbspace Mdm Estimation (MUSUME), obtains multiple low-dimensional spaces that enhance class separability by solving the proposed optimization problem, then the similarity is learned in each low-dimensional space. This multiple projection approach encourages finding the space that is more useful for similarity learning. Experimental evaluation with high-dimensionality EEG datasets (128 channels) confirmed that MUSUME proved significant improvement for classification (p \textless 0.001) and also it showed the potential to beat the existing method relying on only one subspace representation.

4. Riemannian Geometry on Connectivity for Clinical BCI Corsi, Marie-Constance, Chevallier, Sylvain, Yger, Florian, and Noûs, Camille In ICASSP 2021, 2021 [Abs] [HTML] [DOI]

   Riemannian BCI based on EEG covariance have won many data competitions and achieved very high classification results on BCI datasets. To increase the accuracy of BCI systems, we propose an approach grounded on Riemannian geometry that extends this framework to functional connectivity measures. This paper describes the approach submitted to the Clinical BCI Challenge-WCCI2020 and that ranked 1 st on the task 1 of the competition.

5. Purposeful Proposal for CP-afflicted Upper Limbs Exoskeletons Charafeddine, Jinan, Maassarani, Ibrahim, Chevallier, Sylvain, and Alfayad, Samer In 2021 Sixth International Conference on Advances in Biomedical Engineering (ICABME), 2021 [HTML] [DOI]
6. Ensemble learning based on functional connectivity and Riemannian geometry for robust workload estimation Corsi, Marie-Constance, Chevallier, Sylvain, Barthélemy, Quentin, Hoxha, Isabelle, and Yger, Florian In Neuroergonomics conference 2021, 2021 [HTML]
7. Decodable anticipation from prestimulus activity Hoxha, Isabelle, Chevallier, Sylvain, Delorme, Arnaud, Boutin, Arnaud, and Amorim, Michel Ange In 3rd Neuroergonomics Conference, 2021 [HTML]
8. Review of Riemannian Distances and Divergences, Applied to SSVEP-based BCI Chevallier, Sylvain, Kalunga, Emmanuel K., Barthélemy, Quentin, and Monacelli, Eric Neuroinformatics, 2021 [Abs] [HTML] [DOI]

   The firstgeneration of brain-computer interfaces (BCI) classifies multi-channel electroencephalographic (EEG) signals, enhanced by optimized spatial filters.The second generation directly classifies covariance matrices estimated on EEG signals, based on straightforward algorithms such as the minimum-distance-to-Riemannian-mean (MDRM). Classification results vary greatly depending on the chosen Riemannian distance or divergence, whose definitions and reference implementations are spread across a wide mathematical literature. This paper reviews all the Riemannian distances and divergences to process covariance matrices, with an implementation compatible with BCI constraints. The impact of using different metrics is assessed on a steady-state visually evoked potentials (SSVEP) dataset, evaluating centers of classes and classification accuracy. Riemannian approaches embed crucial properties to process EEG data. The Riemannian centers of classes outperform Euclidean ones both in offline and online setups. Some Riemannian distances and divergences have better performances in terms of classification accuracy, while others have appealing computational efficiency.

2020

1. Riemann-Based Algorithms Assessment for Single- and Multiple-Trial P300 Classification in Non-Optimal Environments Delgado, Juan M. Chau, Achanccaray, David, Villota, Elizabeth R., and Chevallier, Sylvain IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2020 [HTML] [DOI]
2. Semi-Supervised Optimal Transport Methods for Detecting Anomalies Alaoui-Belghiti, Amina, Chevallier, Sylvain, Monacelli, Eric, Bao, Guillaume, and Azabou, Eric In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020 [Abs] [HTML] [DOI]

   Building upon advances on optimal transport and anomaly detection, we propose a generalization of an unsupervised and automatic method for detection of significant deviation from reference signals. Unlike most existing approaches for anomaly detection, our method is built on a non-parametric framework exploiting the optimal transportation to estimate deviation from an observed distribution. We described the theoretical background of our method and demonstrate its effectiveness on two datasets: an industrial predictive maintenance task based on audio recording and a detection of anomalous breathing relying on brain signals. In this type of problem, no negative or faulty samples are seen during training and the objective is to detect any abnormal sample without raising false alarm. The proposed approach outperforms all state-of-the-art methods for anomaly detection on the two considered datasets.

3. Geodesically-convex optimization for averaging partially observed covariance matrices Yger, Florian, Chevallier, Sylvain, Barthélemy, Quentin, and Sra, Suvrit In Asian Conference on Machine Learning, 2020 [Abs] [HTML] [DOI]

   Symmetric positive definite (SPD) matrices permeates numerous scientific disciplines, including machine learning, optimization, and signal processing. Equipped with a Riemannian geometry, the space of SPD matrices benefits from compelling properties and its derived Riemannian mean is now the gold standard in some applications, eg brain-computer interfaces (BCI). This paper addresses the problem of averaging covariance matrices with missing variables. This situation often occurs with inexpensive or unreliable sensors, or when artifact-suppression techniques remove corrupted sensors leading to rank deficient matrices, hindering the use of the Riemannian geometry in covariance-based approaches. An alternate but questionable method consists in removing the matrices with missing variables, thus reducing the training set size. We address those limitations and propose a new formulation grounded in geodesic convexity. Our approach is evaluated on generated datasets with a controlled number of missing variables and a known baseline, demonstrating the robustness of the proposed estimator. The practical interest of this approach is assessed on real BCI datasets. Our results show that the proposed average is more robust and better suited for classification than classical data imputation methods.

4. SMR/Theta Neurofeedback Training Improves Cognitive Performance and EEG Activity in Elderly With Mild Cognitive Impairment: A Pilot Study Marlats, Fabienne, Bao, Guillaume, Chevallier, Sylvain, Boubaya, Marouane, Djabelkhir-Jemmi, Leila, Wu, Ya-Huei, Lenoir, Hermine, Rigaud, Anne-Sophie, and Azabou, Eric Frontiers in Aging Neuroscience, 2020 [Abs] [HTML] [DOI]

   Background: Neurofeedback (NF) training, as a method of self-regulation of brain activity, may be beneficial in elderly patients with mild cognitive impairment (MCI). In this pilot study, we investigated whether a sensorimotor (SMR)/theta NF training could improve cognitive performance and brain electrical activity in elderly patients with MCI. Methods: Twenty elderly patients with MCI were assigned to 20 consecutive sessions of sensorimotor (SMR)/theta NF training, during 10 weeks, on a basis of two sessions each week. Neuropsychological assessments and questionnaires, as well as electroencephalogram (EEG), were performed and compared between baseline (T0), after the last NF training session at 10 weeks (T1), and 1-month follow-up (T2). Results: Repeated measures ANOVA revealed that from baseline to post-intervention, participants showed significant improvement in the Montreal cognitive assessment (MoCa, F = 4.78; p = 0.012), the delayed recall of the Rey auditory verbal learning test (RAVLT, F = 3.675; p = 0.032), the Forward digit span (F = 13.82; p \textless 0.0001), the Anxiety Goldberg Scale (F = 4.54; p = 0.015), the Wechsler Adult Intelligence Score–Fourth Edition (WAIS-IV; F = 24.75; p \textless 0.0001), and the Mac Nair score (F = 4.47; p = 0.016). EEG theta power (F = 4.44; p = 0.016) and alpha power (F = 3.84; p = 0.027) during eyes-closed resting-state significantly increased after the NF training and showed sustained improvement at a 1-month follow-up. Conclusion: Our results suggest that NF training could be effective to reduce cognitive deficits in elderly patients with MCI and improve their EEG activity. If these findings are confirmed by randomized controlled studies with larger samples of patients, NF could be seen as a useful non-invasive, non-pharmacological tool for preventing further decline, rehabilitation of cognitive function in the elderly. Clinical Trial Registration: This pilot study was a preliminary step before the trial registered in www.ClinicalTrials.gov, under the number of NCT03526692.

5. Randomized Controlled Study Evaluating Efficiency of Low Intensity Transcranial Direct Current Stimulation (tDCS) for Dyspnea Relief in Mechanically Ventilated COVID-19 Patients in ICU: The tDCS-DYSP-COVID Protocol Azabou, Eric, Bao, Guillaume, Heming, Nicholas, Bounab, Rania, Moine, Pierre, Chevallier, Sylvain, Chevret, Sylvie, Resche-Rigon, Matthieu, Siami, Shidaps, Sharshar, Tarek, Lofaso, Frederic, and Annane, Djillali Frontiers in Medicine, 2020 [Abs] [HTML] [DOI]

   The severe respiratory distress syndrome linked to the new coronavirus disease (COVID-19) includes unbearable dyspneic suffering which contributes to the deterioration of the prognosis of patients in intensive care unit (ICU). Patients are put on mechanical ventilation to reduce respiratory suffering and preserve life. Despite this mechanical ventilation, most patients continue to suffer from dyspnea. Dyspnea is a major source of suffering in intensive care and one of the main factors that affect the prognosis of patients. The development of innovative methods for its management, especially non-drug management is more than necessary. In recent years, numerous studies have shown that transcranial direct current stimulation (tDCS) could modulate the perception of acute or chronic pain. In the other hand, it has been shown that the brain zones activated during pain and dyspnea are close and/or superimposed, suggesting that brain structures involved in the integration of aversive emotional component are shared by these two complex sensory experiences. Therefore, it can be hypothesized that stimulation by tDCS with regard to the areas which, in the case of pain have activated one or more of these brain structures, may also have an effect on dyspnea. In addition, our team recently demonstrated that the application of tDCS on the primary cortical motor area can modulate the excitability of the respiratory neurological pathways. Indeed, tDCS in anodal or cathodal modality reduced the excitability of the diaphragmatic cortico-spinal pathways in healthy subjects. We therefore hypothesized that tDCS could relieve dyspnea in COVID-19 patients under mechanical ventilation in ICU. This study was designed to evaluate effects of two modalities of tDCS (anodal and cathodal) vs. placebo, on the relief of dyspnea in COVID-19 patients requiring mechanical ventilation in ICU.

6. Extending Riemannian Brain-Computer Interface to Functional Connectivity Estimators Chevallier, Sylvain, Corsi, Marie-Constance, Yger, Florian, and Noûs, Camille In IROS Workshop on Bringing geometric methods to robot learning, optimization and control, 2020 [Abs] [HTML]

   This abstract describes a novel approach for handling brain-computer interfaces (BCI), that could be used for robotic applications. State-of-the-art approaches rely on the classification of covariance matrices in the manifold of symmetric positive-definite matrices. Functional connectivity estimators have demonstrated their reliability and are good candidates to improve the classification accuracy of covariance-based methods. This abstract explores possible application of functional connectivity in Riemannian BCI.

2019

1. Neuromotor Strategy of Gait Rehabilitation for Lower-Limb Spasticity Charafeddine, Jinan, Chevallier, Sylvain, Khalil, Mohamad, Pradon, Didier, and Alfayad, Samer In International Conference on Advances in Biomedical Engineering (ICABME), 2019 [Abs] [HTML] [DOI]

   Assisting users and restoring human locomotion for patients with lower limb spasticity is a challenging task. Studies on gait disorders are rare where there is an important variability in movement parameters between patients. Those patients could benefit the most from rehabilitation and assistive mechatronic devices; there is no generic controlling scheme or any dynamical gain indicator. This contribution introduces a bio-kinematic index which is called Neuro-motor index (NMI), based on electromyographic (EMG) and joint angles measurements. NMI is derived from the nonlinear regression with a combination of two co-contraction indices (CCI), which allows addressing the variability of walking situations. This new index is evaluated on patients with cerebral palsy and a stroke. Then, the estimation error was calculated in comparison with the other co-contraction indices. This estimation shows that this index has the highest signification of joint angles prediction. Thus it can be suitable in adaptive rehabilitation control for spasticity cases.

2. Biokinematic Control Strategy for Rehabilitation Exoskeleton Based on User Intention Charafeddine, Jinan, Chevallier, Sylvain, Alfayad, Samer, Khalil, Mohamad, and Pradon, Didier International Journal of Modeling and Optimization, 2019 [Abs] [HTML] [DOI]

   Rehabilitation exoskeletons require a control interface for the direct transfer of mechanical power and exchange of information in order to assist the patient in his/her movements. By using co-contraction indexes (CCI), it is possible to accurately characterize human movement and joint stability. But when dealing with human movement disorders, no existing index allows to achieve neuro-motor control with bio-kinematic sensors. Thus, we propose a neuro-motor interactive method for lower-body exoskeleton control. A novel dynamic index called neuro-motor index (NMI) is introduced to estimate the relation between muscular co-contraction derived from electromyography signals (EMG) and joint angles. To estimate the correlation in the state space and enhance the precision of the NMI, we describe an estimation method relying on a two-way analysis of canonical correlation (CCA). A thorough assessment is presented, by conducting two studies on control subjects and on patients with abnormal gait in a medical environment. i) An offline study on control patients showed that NMI captures the complex variation induced by changing walking speed more accurately than CCI, ii) an online study, applied on successive gait cycles of patients with abnormal walk indicates that the existing CCI have a low accuracy related with joint angles while it is significantly higher with NMI.

3. Des signaux cérébraux aux activités cognitives, approches géométriques et apprentissage statistique Chevallier, Sylvain HDR, Université Paris-Saclay, 2019 [Abs] [HTML]

   Les progrès en analyse des signaux cérébraux pour un décodage en temps réel ont permis des avancées importantes pour les interfaces cérébrales. Ces interfaces cérébrales ou brain-computer interfaces (BCI) permettent de contrôler ou d’échanger des commandes avec un système en nécessitant pas ou peu de capacités physiques. Elles offrent une solution adaptée pour les personnes en situation de handicap. Dans leur forme actuelle, elles requièrent toutefois une bonne capacité de concentration pour fonctionner correctement. Ces BCI ont connu un rapide essor grâce aux méthodes d’apprentissage statistique, mais elles font face à deux verrous scientifiques. Le premier concerne la puissance du signal, qui est très faible, et les bruits environnants qui contaminent les enregistrements de signaux cérébraux, avec des niveaux bien supérieurs à celui du signal d’intérêt. Le second point concerne les variabilités individuelles qui peuvent réduire les algorithmes à des prédictions au niveau de la chance pour 10 à 20}% des sujets. Les travaux présentés dans ce manuscrit abordent les aspects tant expérimentaux que théoriques pour aborder ces deux problèmes. Les approches en traitement du signal utilisées pour travailler sur les signaux cérébraux peuvent dans certains cas s’appliquer sur d’autres types séries temporelles, par exemple pour des applications industrielles sur la détection d’anomalie. Toutes ces approches sont développées avec des outils libres pour une science ouverte et diffusable, en apportant quand c’est possible des contributions aux logiciels libres.

4. Unsupervised anomaly detection using optimal transport for predictive maintenance Alaoui-Belghiti, Amina, Chevallier, Sylvain, and Monacelli, Eric In International Conference on Artificial Neural Networks, 2019 [Abs] [HTML] [DOI]

   Anomaly detection is of crucial importance in industrial environment, especially in the context of predictive maintenance. As it is very costly to add an extra monitoring layer on production machines, non-invasive solutions are favored to watch for precursory clue indicating the possible need for a maintenance operation. Those clues are to be detected in evolving and highly variable working environment, calling for online and unsupervised methods. This contribution proposes a frame- work grounded in optimal transport, for the specific characterization of a system and the automatic detection of abnormal events. This method is evaluated on acoustic dataset and demonstrate the superiority of met- rics derived from optimal transport on the Euclidean ones. The proposed method is shown to outperform one-class SVM on real datasets, which is the state-of-the-art method for anomaly detection.

5. SenseJoy, a pluggable solution for assessing user behavior during powered wheelchair driving tasks Rabreau, Olivier, Chevallier, Sylvain, Chassagne, Luc, and Monacelli, Eric Journal of NeuroEngineering and Rehabilitation, 2019 [Abs] [HTML] [DOI]

   The complex task of Electric Powered Wheelchairs (EPW) prescription relies mainly on personal experience and subjective observations despite standardized processes and protocols. The most informative measurements come from joystick monitoring, but recording direct joystick outputs require to disassemble the joystick. We propose a new solution called "SenseJoy" that is easy to plug on a joystick and is suitable to characterize the driver behavior by estimating the joystick command. Methods: SenseJoy is a pluggable system embedded on EPW built with a 3D accelerometer and a 2D gyrometer placed within the joystick and another 3D accelerometer located at the basis of the joystick. Data is sampled at 39 Hz and processed offline. First, SenseJoy sensitivity is assessed on wheelchair driving tasks performed by a group of 8 drivers (31 ± 8 years old, including one driver with left hemiplegia, one with cerebral palsy) in a lab environment. Direct joystick measurements are compared with SenseJoy estimations in different driving exercises. A second group of 5 drivers is recorded in the ecological context of a rehabilitation center (41 ± 10 years old, with two tetraplegic drivers, one tetraplegic driver with cognitive disorder, one driver post-stroke, one driver with right hemiplegia). The measurements from all groups of drivers are evaluated with an unsupervised statistical analysis, to estimate driving profile clusters. Results: The SenseJoy is able to measure the EPW joystick inclination angles with a resolution of 1.31% and 1.23% in backward/forward and left/right directions respectively. A statistical validation ensures that the classical joystick-based indicators are equivalent when acquired with the SenseJoy or with a direct joystick output connection. Using an unsupervised methodology, based on a similarity matrix between subjects, it is possible to characterize the driver profile from real data. Conclusion: SenseJoy is a pluggable system for assessing the joystick controls during EPW driving tasks. This system can be plugged on any EPW equipped with a joystick control interface. We demonstrate that it correctly estimates the performance indicators and it is able to characterize driving profile. The system is suitable and efficient to assist therapists in their recommendation, by providing objective measures with a fast installation process.

6. Toward bio-kinematic for secure use of rehabilitation exoskeleton Charafeddine, Jinan, Pradon, Didier, Alfayad, Samer, Chevallier, Sylvain, and Khalil, Mohamad In Computer Methods in Biomechanics and Biomedical Engineering, 2019 [HTML] [DOI]

2018

1. Riemannian classification for SSVEP based BCI: offline versus online implementations Chevallier, Sylvain, Kalunga, Emmanuel K., Barthélemy, Quentin, and Yger, Florian In Brain Computer Interfaces Handbook: Technological and Theoretical Advances, CRC Press, 2018 [Abs] [HTML]

   This chapter focus on the different implementations of brain-computer interface (BCI) based on Steady-State Visually Evoked Potentials (SSVEPs). In offline BCI, feature extraction and the classification are performed at the end of the session, when all trials are available. Whereas, in on-line settings, they are performed several times during each trial, usually for each available epoch recorded by the electroencephalogram (EEG) device, enabling real-time and asynchronous BCI. A recent successful approach in feature extraction and signal processing for BCI is Riemannian geometry , which deals with covariance matrices. They capture the degree of correlation between several random variables, that is how the brain signals change relatively to each other. These techniques have demonstrated their benefit on several occasions, leading to winning algorithms in international competitions and to state-of-the-art results on renowned BCI benchmarks. After reviewing some of the most robust approaches in feature extraction for SSVEP, this chapter will introduce newer tools based on Riemannian geometry. With an application to SSVEP, this article shows through a comparison how Riemannian geometry allows one to easily define offline and online implementations that have better accuracies than state of the art.

2. Transfer learning for SSVEP-based BCI using Riemannian similarities between users Kalunga, Emmanuel K, Chevallier, Sylvain, and Barthélemy, Quentin In 2018 26th European Signal Processing Conference (EUSIPCO), 2018 [Abs] [HTML]

   Brain-Computer Interfaces (BCI) face a great challenge: how to harness the wide variability of brain signals from a user to another. The most visible problem is the lack of a sound framework to capture the specificity of a user brain waves. A first attempt to leverage this issue is to design user-specific spatial filters, carefully adjusted with a lengthy calibration phase. A second, more recent, opening is the systematic study of brain signals through their covariance, in an appropriate space from a geometric point of view. Riemannian geometry allows to efficiently characterize the variability of inter-subject EEG, even with noisy or scarce data. This contribution is the first attempt for SSVEP-based BCI to make the most of the available data from a user, relying on Riemannian geometry to estimate the similarity with a multi-user dataset. The proposed method is built in the framework of transfer learning and borrows the notion of composite mean to partition the space. This method is evaluated on 12 subjects performing an SSVEP task for the control of an exoskeleton arm and the results show the contribution of Riemannian geometry and of the user-specific composite mean, whereas there is only a few data available for a subject.

3. Supra-threshold Inspiratory Loads Elicit Respiratory Related Evoked Potentials in Healthy Subjects Azabou, Eric, Bao, Guillaume, Chevallier, Sylvain, Marlats, Fabienne, Brussel, Bernard, Mayaud, Louis, Prigent, Hélène, Petitjean, Michel, and Lofaso, Frédéric In Neurophysiologie Clinique/Clinical Neurophysiology, 2018 [Abs] [HTML] [DOI]

   Introduction Respiratory-related evoked potential (RREP) are well established cerebral correlates of respiratory perceptions. The early components N1/P1 reflect first-order sensory processing of afferent respiratory signals in brain sensorimotor regions while the late ones P2/P3 involve higher-order cognitive processing [1]. The current stimuli used to elicited RREP are short occlusions at the onset of an inspiration or a transiant interruption of the inspiration [1]. Durations of the inspiratory occlusion do not affect RREP components [2] while there is a linear relationship between inspiratory loads and RREP components amplitudes. Moreover, an inspiration load need to reach some threshold value to elicit RREP [1,3]. Here, we investigated whether supra-threshold load inspiration could induce RREP without occlusion. Methods: EEG was synchronously recorded with the respiratory signals in 11 healthy volunteers breathing into a mouthpiece connected to a non-rebreathing valve. Four consecutives inspiration loads were produced by a pressure-threshold device connected to the non-rebreathing valve: no_Load, 10, 20 and 30 cmH2O/L/sec. Each load was presented during 5 minutes and there is a 3-min resting time between two consecutives loads presentations. Magnitude estimation (ME) for each load was scored with the A1 subscale of the Multi-dimensional profile of dyspnea. Inspiration onset time locked EEG epochs were averaged for each breathing conditions and examined for presence of RREP components. Results: The no_load condition did not elicit any peak of RREP, but 10, 20 and 30 cmH2O/L/sec produced N1 and P3 RREP components with amplitudes correlating with the loads as well as the loads’ magnitude estimation. Mean amplitudes of the N1 and P3 RREP components for 10, 20 and 30 cmH2O/L/sec were respectively 3.6 ±0.5, 5.6 ±0.9, 3.8 ±1.3 and-4.5 ±1.3,-11.4 ±2.4,-7.6 ±2.2 µV. Conclusion: Supra-threshold load inspiration could induce RREP without occlusion. This method should be compared with the standard inspiration occlusion method.

4. Brain-Machine Interface for Mechanical Ventilation Using Respiratory-Related Evoked Potential Chevallier, Sylvain, Bao, Guillaume, Hammami, Mayssa, Marlats, Fabienne, Mayaud, Louis, Annane, Djillali, Lofaso, Frédéric, and Azabou, Eric In International Conference on Artificial Neural Networks (ICANN), 2018 [Abs] [HTML] [DOI]

   The correct ventilation for patients in intensive care units plays a critical role for the prognostic and the recovery during the stay in the hospital. Desynchronization between the ventilator and the patient is an important source of stress, emphasized by the lack of communication due to intubation or loss of consciousness. This contribution proposes a novel approach based on electroencephalographic (EEG) activity to detect breathing effort. Relying both on recent neuroscience finding on respiratory-related evoked potential and on latest development of information geometry, the proposed approach elaborates on Rieman-nian distances between EEG covariance matrices to differentiate among different respiratory loads. The results demonstrate that this approach outperform existing state-of-the-art methods quantitatively, in terms of mean accuracy, and qualitatively, being able to predict level of breathing discomfort.

2017

1. Inductive Means and Sequences Applied to Online Classification of EEG Massart, Estelle M, and Chevallier, Sylvain In International Conference on Geometric Science of Information, 2017 [Abs] [HTML] [DOI]

   The translation of brain activity into user command, through Brain-Computer Interfaces (BCI), is a very active topic in machine learning and signal processing. As commercial applications and out-of-the-lab solutions are proposed, there is an increased pressure to provide online algorithms and real-time implementations. Electroencephalography (EEG) systems offer lightweight and wearable solutions, at the expense of signal quality. Approaches based on covariance matrices have demonstrated good robustness to noise and provide a suitable representation for classification tasks, relying on advances in Riemannian geometry. We propose to equip the minimum distance to mean (MDM) classifier with a new family of means, based on the inductive mean, for block-online classification tasks and to embed the inductive mean in an incremental learning algorithm for online classification of EEG.

2. Subject-specific time-frequency selection for multi-class motor imagery-based BCIs using few Laplacian EEG channels Yang, Yuan, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle Biomedical Signal Processing and Control, 2017 [Abs] [HTML] [DOI]

   The essential task of a motor imagery brain–computer interface (BCI) is to extract the motor imagery-related features from electroencephalogram (EEG) signals for classifying motor intentions. However, the optimal frequency band and time segment for extracting such features differ from subject to subject. In this work, we aim to improve the multi-class classification and to reduce the required EEG channel in motor imagery-based BCI by subject-specific time-frequency selection. Our method is based on a criterion namely Fisher discriminant analysis-type F-score to simultaneously select the optimal frequency band and time segment for multi-class classification. The proposed method uses only few Laplacian EEG channels (C3, Cz and C4) located around the sensorimotor area for classification. Applied to a standard multi-class BCI dataset (BCI competition III dataset IIIa), our method leads to better classification performance and smaller standard deviation across subjects compared to the state-of-art methods. Moreover , adding artifacts contaminated trials to the training dataset does not necessarily deteriorate our classification results, indicating that our method is tolerant to artifacts.

2016

1. Online SSVEP-based BCI using Riemannian geometry Kalunga, Emmanuel K., Chevallier, Sylvain, Barthélemy, Quentin, Djouani, Karim, Monacelli, Eric, and Hamam, Yskandar Neurocomputing, 2016 [Abs] [HTML] [DOI]

   Challenges for the next generation of Brain Computer Interfaces (BCI) are to mitigate the common sources of variability (electronic, electrical, biological) and to develop online and adaptive systems following the evolution of the subject׳s brain waves. Studying electroencephalographic (EEG) signals from their associated covariance matrices allows the construction of a representation which is invariant to extrinsic perturbations. As covariance matrices should be estimated, this paper first presents a thorough study of all estimators conducted on real EEG recording. Working in Euclidean space with covariance matrices is known to be error-prone, one might take advantage of algorithmic advances in Riemannian geometry and matrix manifold to implement methods for Symmetric Positive-Definite (SPD) matrices. Nonetheless, existing classification algorithms in Riemannian spaces are designed for offline analysis. We propose a novel algorithm for online and asynchronous processing of brain signals, borrowing principles from semi-unsupervised approaches and following a dynamic stopping scheme to provide a prediction as soon as possible. The assessment is conducted on real EEG recording: this is the first study on Steady-State Visually Evoked Potential (SSVEP) experimentations to exploit online classification based on Riemannian geometry. The proposed online algorithm is evaluated and compared with state-of-the-art SSVEP methods, which are based on Canonical Correlation Analysis (CCA). It is shown to improve both the classification accuracy and the information transfer rate in the online and asynchronous setup.

2. Smart Interfaces for New Building Design Process Martin, Hugo, Chevallier, Sylvain, and Monacelli, Eric In IEEE Last Mile Smart Mobility, 2016 [Abs] [HTML]

   Building information modeling opened new hori- zons for designing future urban spaces. Logistics and mobility issues are difficult to take into account during the construction process, as they require a precise global vision and the access to information yields by different building departments (architecture, methods, structure...). This contribution proposes a novel approach, based on a collaborative framework, to develop the building model, that is to ensure the quality of logistic and mobility during construction, with a time-efficient methodology.

3. Embedded-BCI: assessment of parallelizing computations on an embedded system Zammouri, Amin, and Chevallier, Sylvain In IEEE Last Mile Smart Mobility, 2016 [Abs] [HTML]

   Using Brain-Computer Interfaces (BCI) as an assistive technology aims at providing an innovative solution adapted to subjects’ disabilities. BCI either provide a new interface for controlling solution mobility (e.g. wheelchair) or monitoring the state of user during his/her journey. This would be possible by implementing these interfaces on Embedded Systems (ES). However, because of the BCI sophisticated data processing and the ES limited computation performances, the computation time for a real-time use of the BCI on an ES is a limitation. Hence in this work, we investigate and evaluate the parallelization and acceleration performances, on a Raspberry Pi 2 model B (RPi) board, of an STFT-based algorithm for estimating cognitive workload from an Electroencephalographic (EEG) signal. This is done based on multi-core CPU and GPU architectures of the used RPi. Results show that the parallelized implementation using the CPU runs up to × 4 faster than a simple implementation. Compared to CPU of intel-CORE i3 processor, the GPU of the RPi revealed large difference in computation time.

4. Subject-Specific Channel Selection Using Time Information for Motor Imagery Brain–Computer Interfaces Yang, Yuan, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle Cognitive Computation, 2016 [Abs] [HTML] [DOI]

   Keeping a minimal number of channels is essential for designing a portable brain–computer interface system for daily usage. Most existing methods choose key channels based on spatial information without optimization of time segment for classification. This paper proposes a novel subject-specific channel selection method based on a criterion called F score to realize the parameterization of both time segment and channel positions. The F score is a novel simplified measure derived from Fisher’s discrimi-nant analysis for evaluating the discriminative power of a group of features. The experimental results on a standard dataset (BCI competition III dataset IVa) show that our method can efficiently reduce the number of channels (from 118 channels to 9 in average) without a decrease in mean classification accuracy. Compared to two state-of-the-art methods in channel selection, our method leads to comparable or even better classification results with less selected channels.

2015

1. From Euclidean to Riemannian Means: Information Geometry for SSVEP Classification Kalunga, Emmanuel K., Chevallier, Sylvain, Barthélemy, Quentin, Djouani, Karim, Hamam, Yskandar, and Monacelli, Eric In Geometric Science of Information, 2015 [HTML]
2. Intelligent ocular artifacts removal in a noninvasive singlechannel EEG recording Zammouri, Amin, Ait Moussa, Abdelaziz, Chevallier, Sylvain, and Monacelli, Eric In Intelligent Systems and Computer Vision (ISCV), 2015 [Abs] [HTML] [DOI]

   Muscle noises, line noises and eye movements are the main interferences that make difficulties when interpreting and analyzing electroencephalographic signals. Many methods have been proposed for artifacts removing from EEG measurements, and especially those arising from an ocular source. Principal Component Analysis (PCA) and Independent Component Analysis (ICA) have been proposed to remove ocular artifacts from multichannel EEG. In contrast to this, we present a new algorithm for ocular artifacts removal from a single electroencephalographic channel recording. This method is based on a set of information on brain wave frequencies. Our results on EEG data, collected from healthy subjects, show that our algorithm can effectively detect and remove ocular artifacts in EEG recordings.

3. Adaptive visualization system for construction building information using saliency Martin, Hugo, Chevallier, Sylvain, and Monacelli, Eric In International Conference on Construction Applications of Virtual Reality (CONVR), 2015 [Abs] [HTML]

   Building Information Modeling (BIM) is a recent construction process based on a 3D model, containing every component related to the building achievement. Architects, structure engineers, method engineers, and others departments to the building process work in collaboration on this model through the design-to-construction cycle. The high complexity and the large amount of information driven by novelization in the model raise several issues and delaying its wide adoption in the industrial world. One of the most important is the visualization: professionals have difficulties to find out the relevant information for their job. Actual solutions suffer from two limitations: the BIM models information are processed manually and insignificant information are simply hidden, leading to inconsistencies in the building model. This paper describes a system relying on an ontological representation of the building information to label automatically the building elements. Depending on the user’s department, the visualization is modified according to these labels by automatically adjusting the colors and image properties based on a saliency model. The proposed saliency model incorporates several adaptations to fit the specificities of architectural images.

4. Data augmentation in Riemannian space for Brain-Computer Interfaces Kalunga, Emmanuel K., Chevallier, Sylvain, and Barthélemy, Quentin In ICML Workshop on Statistics, Machine Learning and Neuroscience (Stamlins 2015), 2015 [Abs] [HTML]

   Brain-Computer Interfaces (BCI) try to interpret brain signals , such as EEG , to issue some command or to characterize the cognitive states of the subjects. A strong limitation is that BCI tasks require a high concentration of the user , de facto limiting the length of experiment and the size of the dataset. Furthermore , several BCI paradigms depend on rare events , as for event-related potentials , also reducing the number of training examples available. A common strategy in machine learning when dealing with scarce data is called data augmentation ; new samples are generated by applying chosen transformations on the original dataset. In this contribution , we propose a scheme to adapt data augmentation in EEG-based BCI with a Riemannian standpoint : geometrical properties of EEG covariance matrix are taken into account to generate new training samples. Neural network are good candidates to benefit from such training scheme and a simple multi-layer perceptron offers good results . Experimental validation is conducted on two datasets : an SSVEP experiment with few training samples in each class and an error potential experiment with unbalanced classes (NER Kaggle competition).

5. Using Riemannian geometry for SSVEP-based Brain Computer Interface Kalunga, Emmanuel K., Chevallier, Sylvain, and Barthélemy, Quentin In arXiv, 2015 [HTML]

2014

1. Artificial Neurogenesis: An Introduction and Selective Review Kowaliw, Taras, Bredeche, Nicolas, Chevallier, Sylvain, and Doursat, René In Growing Adaptive Machines, Springer, 2014 [Abs] [HTML]

   In this introduction and review—like in the book which follows—we explore the hypothesis that adaptive growth is a means of producing brain-like machines. The emulation of neural development can incorporate desirable characteristics of natural neural systems into engineered designs. The introduction begins with a review of neural development and neural models. Next, artificial development— the use of a developmentally-inspired stage in engineering design—is introduced. Several strategies for performing this " meta-design " for artificial neural systems are reviewed. This work is divided into three main categories: bio-inspired representations ; developmental systems; and epigenetic simulations. Several specific network biases and their benefits to neural network design are identified in these contexts. In particular, several recent studies show a strong synergy, sometimes interchange-ability, between developmental and epigenetic processes—a topic that has remained largely under-explored in the literature.

2. Time-frequency optimization for discrimination between imagination of right and left hand movements based on two bipolar electroencephalography channels Yang, Yuan, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle EURASIP Journal on Advances in Signal Processing, 2014 [Abs] [HTML] [DOI]

   To enforce a widespread use of efficient and easy to use brain-computer interfaces (BCIs), the inter-subject robustness should be increased and the number of electrodes should be reduced. These two key issues are addressed in this contribution, proposing a novel method to identify subject-specific time-frequency characteristics with a minimal number of electrodes. In this method, two alternative criteria, time-frequency discrimination factor (TFDF) and F score, are proposed to evaluate the discriminative power of time-frequency regions. Distinct from classical measures (e.g., Fisher criterion, r 2 coefficient), the TFDF is based on the neurophysiologic phenomena, on which the motor imagery BCI paradigm relies, rather than only from statistics. F score is based on the popular Fisher’s discriminant and purely data driven; however, it differs from traditional measures since it provides a simple and effective measure for quantifying the discriminative power of a multi-dimensional feature vector. The proposed method is tested on BCI competition IV datasets IIa and IIb for discriminating right and left hand motor imagery. Compared to state-of-the-art methods, our method based on both criteria led to comparable or even better classification results, while using fewer electrodes (i.e., only two bipolar channels, C3 and C4). This work indicates that time-frequency optimization can not only improve the classification performance but also contribute to reducing the number of electrodes required in motor imagery BCIs.

3. Influence of extracellular oscillations on neural communication: a computational perspective Tiganj, Zoran, Chevallier, Sylvain, and Monacelli, Eric Frontiers in Computational Neuroscience, 2014 [Abs] [HTML] [DOI]

   Neural communication generates oscillations of electric potential in the extracellular medium. In feedback, these oscillations affect the electrochemical processes within the neurons, influencing the timing and the number of action potentials. It is unclear whether this influence should be considered only as noise or it has some functional role in neural communication. Through computer simulations we investigated the effect of various sinusoidal extracellular oscillations on the timing and number of action potentials. Each simulation is based on a multicompartment model of a single neuron, which is stimulated through spatially distributed synaptic activations. A thorough analysis is conducted on a large number of simulations with different models of CA3 and CA1 pyramidal neurons which are modeled using realistic morphologies and active ion conductances. We demonstrated that the influence of the weak extracellular oscillations, which are commonly present in the brain, is rather stochastic and modest. We found that the stronger fields, which are spontaneously present in the brain only in some particular cases (e.g., during seizures) or that can be induced externally, could significantly modulate spike timings.

4. On the need for metrics in dictionary learning assessment Chevallier, Sylvain, Barthélemy, Quentin, and Atif, Jamal In European Signal Processing Conference (EUSIPCO), 2014 [Abs] [HTML]

   Dictionary-based approaches are the focus of a growing attention in the signal processing community, often achieving state of the art results in several application fields. Albeit their success , the criteria introduced so far for the assessment of their performances suffer from several shortcomings. The scope of this paper is to conduct a thorough analysis of these criteria and to highlight the need for principled criteria, enjoying the properties of metrics. Henceforth we introduce new criteria based on transportation like metrics and discuss their behaviors w.r.t the literature.

5. Subspace metrics for multivariate dictionaries and application to EEG Chevallier, Sylvain, Barthélemy, Quentin, and Atif, Jamal In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2014 [Abs] [HTML] [DOI]

   Overcomplete representations and dictionary learning algorithms are attracting a growing interest in the machine learning community. This paper addresses the emerging problem of comparing multivari-ate overcomplete dictionaries. Despite a recurrent need to rely on a distance for learning or assessing multivariate overcomplete dictionaries , no metrics in their underlying spaces have yet been proposed. Henceforth we propose to study overcomplete representations from the perspective of matrix manifolds. We consider distances between multivariate dictionaries as distances between their spans which reveal to be elements of a Grassmannian manifold. We introduce set-metrics defined on Grassmannian spaces and study their properties both theoretically and numerically. Thanks to the introduced met-rics, experimental convergences of dictionary learning algorithms are assessed on synthetic datasets. Set-metrics are embedded in a clustering algorithm for a qualitative analysis of real EEG signals for Brain-Computer Interfaces (BCI). The obtained clusters of subjects are associated with subject performances. This is a major method-ological advance to understand the BCI-inefficiency phenomenon and to predict the ability of a user to interact with a BCI.

6. Hybrid interface: Integrating BCI in multimodal human-machine interfaces Kalunga, Emmanuel K., Chevallier, Sylvain, Rabreau, Olivier, and Monacelli, Eric In IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), 2014 [Abs] [HTML] [DOI]

   In the context of assistive technologies, it is important to design systems that adapt to the user specificities, and to rely as much as possible on the residual capacities of each user. We define a new methodology in the context of assistive robotics: it is an hybrid approach where a physical interface is complemented by a Brain-Computer Interface (BCI). An implementation of such methodology is proposed, using a 3D touchless interface for continuous control and a steady-state visually evoked potential (SSVEP)-based BCI for triggering specific actions. We describe a novel algorithm for classification of SSVEP signals based on Canonical Correlation Analysis (CCA) and Support Vector Machines (SVM). Its reliability and robustness are assessed in an online setup and its results are compared to existing algorithms. Finally, an experimental evaluation of the proposed system is performed with a 3D navigation task in a Virtual Environment (VE). The system is also embedded on an assistive robotic arm exoskeleton to validate its feasibility.

2013

1. SSVEP enhancement based on Canonical Correlation Analysis to improve BCI performances Kalunga, Emmanuel K., Djouani, Karim, Hamam, Yskandar, Chevallier, Sylvain, and Monacelli, Eric In AFRICON, 2013 [Abs] [HTML] [DOI]

   Brain Computer Interfaces (BCI) rely on brain waves signal, such as electro-encephalogram (EEG) recording, to endow a disabled user with non-muscular communication. Given the very low signal-to-noise ratio of EEG, a signal enhancement phase is crucial for ensuring decent performances in BCI systems. Several methods have been proposed for EEG signal enhancement, such as Independent Component Analysis, Common Spatial Pattern, and Principal Component Analysis. We show that Canonical Correlation Analysis (CCA), initially introduced to SSVEP-based BCI as a feature extraction method, is a good candidate for such preprocessing state. Evaluation is performed on a recording from 5 subjects during a BCI task based on Steady-State Visual Evoked Potentials (SSVEP). The authors demonstrate that CCA significantly improves classification performances in SSVEP-based BCIs.

2012

1. A self-paced hybrid BCI based on EEG and EOG Yang, Yuan, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle In Workshop of Tools for Brain-Computer Interaction (TOBI), 2012 [HTML]
2. Automatic selection of the number of spatial filters for motor-imagery BCI Yang, Yuan, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle In European Symposium on Artificial Neural Networks (ESANN), 2012 [Abs] [HTML]

   Common Spatial Pattern (CSP) is widely used for constructing spatial filters to extract features for motor-imagery-based BCI. One main parameter in CSP-based classification is the number of spatial filters used. An automatic method relying on Rayleigh quotient is presented to estimate its optimal value for each subject. Based on an existing dataset, we validate the contribution of the proposed method through a study of the effect of this parameter on the classification performance. The evaluation on testing data shows that the estimated subject-specific optimal values yield better performances than the recommanded value in the literature.

3. Time-frequency Selection in Two Bipolar Channels for Improving the Classification of Motor Imagery EEG Yang, Yang, Chevallier, Sylvain, Wiart, Joe, and Bloch, Isabelle In Engineering in Medicine and Biology Conference (EMBC), 2012 [HTML]
4. Online frequency band estimation and change-point detection Tiganj, Zoran, Mboup, Mamadou, Chevallier, Sylvain, and Kalunga, Emmanuel Systems and Computer Science (ICSCS), 2012 1st International Conference on, 2012 [HTML]
5. Fast calibration of hand movement-based interface for arm exoskeleton control Martin, Hugo, Chevallier, Sylvain, and Monacelli, Eric In European Symposium on Artificial Neural Networks (ESANN), 2012 [Abs] [HTML]

   Several muscular degenerative diseases alter motor abilities of large muscles but spare smaller muscles, e.g. keeping hand motor skills relatively unaffected while upper limbs ones are altered. Thus, hand movements could be be used to control an arm exoskeleton for rehabilitation and assistive purpose. Using an infra-red sensors (IR) based interface for the exoskeleton control, this paper describes the learning part of the system, endowing the system with a fast online calibration and adaptation abilities. This learning component shows good results and have been successfully implemented on the real system.

2011

1. An Introduction to Deep Learning Arnold, Ludovic, Rebecchi, Sebastien, Chevallier, Sylvain, and Paugam-Moisy, Hélène In European Symposium on Artificial Neural Networks (ESANN), 2011 [Abs] [HTML]

   The deep learning paradigm tackles problems on which shallow architectures (e.g. SVM) are affected by the curse of dimensionality. As part of a two-stage learning scheme involving multiple layers of non-linear processing a set of statistically robust features is automatically extracted from the data. The present tutorial introducing the ESANN deep learning special session details the state-of-the-art models and summarizes the current understanding of this learning approach which is a reference for many difficult classification tasks.

2. Emergence of Temporal and Spatial Synchronous Behaviors in a Foraging Swarm Chevallier, Sylvain, Bredeche, Nicolas, Paugam-Moisy, Hélène, and Sebag, Michèle In European Conference on Artificial Life, 2011 [Abs] [HTML]

   Biological populations often exhibit complex and efficient behaviors, where temporal and spatial couplings at the macro-scale population level emerge from interactions at the micro-scale individual level, without any centralized control. This paper specifically investigates the emergence of behavioral synchronization and the division of labor in a foraging swarm of robotic agents. A deterministic model is proposed and used by each agent to decide whether it goes foraging, based on local cues about its fellow ants’ behavior. This individual model, based on the competition of two spiking neurons, results in a self-organized division of labor at the population level. Depending on the strength and occurrences of interactions among individuals, the population behavior displays either an asynchronous, or a synchronous aperiodic, or a synchronous periodic division of labor. Further, the benefits of synchronized individual behaviors in terms of overall foraging efficiency are highlighted in a 2D spatial simulation.

2010

1. SpikeAnts, un réseau de neurones impulsionnels pour modéliser l’organisation émergente dans un système complexe Chevallier, Sylvain, Paugam-Moisy, Hélène, and Sebag, Michèle In Proc. of the 5th french conference on Computational Neuroscience, 2010 [HTML]
2. Efficient neural models for visual attention Chevallier, Sylvain, Cuperlier, Nicolas, and Gaussier, Philippe In International Conference on Computer Vision and Graphics, 2010 [Abs] [HTML]

   Human vision rely on attention to select only a few regions to process and thus reduce the complexity and the processing time of visual task. Artificial vision systems can benefit from a bio-inspired attentional process relying on neural models. In such applications, what is the most efficient neural model: spiked-based or frequency-based? We propose an evaluation of both neural model, in term of complexity and quality of results (on artificial and natural images).

3. SpikeAnts, a spiking neuron network modelling the emergence of organization in a complex system Chevallier, Sylvain, Paugam-Moisy, Hélène, and Sebag, Michèle In Advances in Neural Information Processing Systems 23, 2010 [HTML]

2009

1. Implementation of a preattentional system with spiking neurons Chevallier, Sylvain Thèse, Université Paris-Sud, 2009 [Abs] [HTML]

   Spiking neurons seem to capture important characteristics of biological neurons with relatively simple models. Their main interest is their ability to code information as discrete events. We have investigated the possible contributions of these models to artficial vision. Due to vision constraints, particularly for processing time, we use simple spiking neural models. A spiking neuron network architecture is described for encoding and extracting saliencies, thanks to the discretization induced by spiking neurons. The saliency map is a combination, in space and temporal domain, of visual modality maps (contrasts, orientations and colors) for different spatial scales. We propose a neural filtering method for building the modality maps. This method implements a process, similar to convolution filtering, by gradual approximation : the more processing time is allowed to the algorithm, the better is the approximation. The architecture ranks the saliencies in the order of their interest, with the most important encoded first. Another spiking neuron network, inspired by dynamic neural fields, is attached on top of the architecture, allowing to select the most salient region and to focus on. Experimental results point out that the architecture is able to extract saliencies in a sequence of images, to select the most salient region and to focus on this region, in presence of noise or when the saliency moves.

2. Difference of Gaussians type neural image filtering with spiking neurons Chevallier, Sylvain, and Dahdouh, Sonia In International Conference on Neural Computation, 2009 [Abs] [HTML]

   This contribution describes a bio-inspired image filtering method using spiking neurons. Bio-inspired approaches aim at identifying key properties of biological systems or models and proposing efficient implementations of these properties. The neural image filtering method takes advantage of the temporal integration behavior of spiking neurons. Two experimental validations are conducted to demonstrate the interests of this neural-based method. The first set of experiments compares the noise resistance of a classical DOG filtering method and the neuronal DOG method on a synthetic image. The other experiment explores the edges recovery ability on a natural image. The results show that the neural-based DOG filtering method is more resistant to noise and provides a better edge preservation than classical DOG filtering method.

2008

1. Implémentation d’un mécanisme de “covert attention” avec un réseau de neurones impulsionnels Chevallier, Sylvain, and Tarroux, Philippe In NeuroComp 2e conférence française de Neurosciences Computationnelles, 2008 [HTML]
2. Visual focus with spiking neurons Chevallier, Sylvain, and Tarroux, Philippe In European Symposium on Artificial Neural Networks (ESANN), 2008 [Abs] [HTML]

   We propose to implement a network of leaky integrate-and-fire neurons able to detect and to focus on a stimulus even in the presence distractors. The experimental data show that this behavior is very robust to noise. This process is similar to an early visual attention mechanism.

3. Covert attention with a spiking neural network Chevallier, Sylvain, and Tarroux, Philippe In Int. Conf. on Computer Vision Systems (ICVS), 2008 [HTML]

2006

1. Extraction de saillances par un réseau de neurones impulsionnels Chevallier, Sylvain, and Tarroux, Philippe In NeuroComp 1ère conférence française de Neurosciences Computationnelles, 2006 [HTML]
2. Saliency extraction with a distributed spiking neural network Chevallier, Sylvain, Tarroux, Philippe, and Paugam-Moisy, Hélène In European Symposium on Artificial Neural Networks (ESANN), 2006 [HTML]

2005

1. Vers une architecture fonctionnelle et distribuée en vision artificielle Chevallier, S. Rapport de Master, Orsay, France, 2005 [HTML]
2. Distributed processing for modelling real-time multimodal perception in a virtual robot Chevallier, Sylvain, Paugam-Moisy, Hélène, and Lemaitre, François In Int. Conf. on Parallel and Distributed Computing and Networks (PDCN), 2005 [HTML]
3. Modélisation de processus attentionnels dans la perception multimodale d’un robot virtuel Chevallier, Sylvain, and Paugam-Moisy, Hélène In Actes du VIème Colloque Jeunes Chercheurs en Sciences Cognitives, 2005 [HTML]