Article
Neurosciences
Boris Yazmir, Miriam Reiner
Summary: The manipulation of remote agents in remote surgery or BCI-wheelchair control often results in errors, particularly related to user intent misclassification and interface system errors. This study focuses on errors caused by unpredicted interface movements that violate user intent and lead to sensory conflicts. The effects of congruent and incongruent sensory stimuli induced by interface errors are examined, with a specific focus on haptic and visual cues. The study identifies the prototypical patterns of EEG error signals associated with these two types of interface errors and reveals significant differences in EEG potentials between them.
Article
Clinical Neurology
Una Pale, Tomas Teijeiro, David Atienza
Summary: Long-term monitoring of patients with epilepsy is a challenging problem in engineering. This study proposes a novel semi-supervised learning approach based on multi-centroid HD computing, which shows significantly improved performance in epilepsy detection, especially in cases of data imbalance.
FRONTIERS IN NEUROLOGY
(2022)
Article
Construction & Building Technology
JungHo Jeon, Hubo Cai
Summary: This study investigates the feasibility of identifying construction hazards by developing an EEG classifier based on experiments conducted in an immersive virtual reality environment. The CatBoost classifier achieved the highest performance with 95.1% accuracy, while also identifying key channel locations and frequency bands closely associated with hazard perception.
AUTOMATION IN CONSTRUCTION
(2021)
Article
Computer Science, Information Systems
Rishikanth Chandrasekaran, Fatemeh Asgareinjad, Justin Morris, Tajana Rosing
Summary: This research paper introduces three approaches to multi-label classification, considering the complexity of the problem and striking a balance between computational efficiency and accuracy. The methods demonstrate a significant improvement in computational efficiency while maintaining comparable accuracy.
Article
Engineering, Biomedical
Alisha Menon, Daniel Sun, Sarina Sabouri, Kyoungtae Lee, Melvin Aristio, Harrison Liew, Jan M. Rabaey
Summary: The proposed hyperdimensional computing (HDC) architecture achieves significant energy efficiency improvement in the biomedical field and outperforms traditional support vector machine (SVM) processors in terms of energy efficiency.
IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS
(2022)
Article
Neurosciences
Bo Mu, Chang Niu, Jingping Shi, Rumei Li, Chao Yu, Kuiying Yin
Summary: The experiment successfully discovered the presence of ErrP in cerebellar regions and developed an effective feature extraction method for EEG data. By extracting time-frequency features and using SBS selection to screen channels, we achieved an average accuracy of 80.5% for two-class classification of executive segments in the EEG data.
Article
Clinical Neurology
Kaspar A. A. Schindler, Abbas Rahimi
Summary: The under-sampling of disease dynamics in epilepsy patients in the typical clinical and EEG examination setting is a central challenge today. Implantable devices for monitoring brain signals and detecting seizures could significantly improve this situation and inform personalized treatment at an unprecedented scale. These devices should be optimized for energy efficiency and compact design to improve maintenance and battery life.
FRONTIERS IN NEUROLOGY
(2021)
Article
Psychology, Biological
Victor J. Pokorny, Scott R. Sponheim, Eric Rawls
Summary: This study examined the effect of reduced-dimensionality ICA in cleaning EEG data and found that PCA-based rdICA had a significant impact on the mean amplitude of early sensory components under certain conditions, but the impact on other aspects was inconsistent between datasets.
Article
Psychology, Biological
Hans Revers, Katrijn Van Deun, Jean Vroomen, Marcel Bastiaansen
Summary: This study investigated the neural correlates of facial attractiveness using EEG recordings while presenting pictures of male or female faces with different attractiveness levels to participants. Results showed salience and valence effects in the EEG patterns, indicated by ERP components and confirmed by MVPA classifications. The effects were only observed for the preferred gender faces, suggesting that facial attractiveness elicits neural responses related to valenced experiences but only if the faces are considered relevant.
BIOLOGICAL PSYCHOLOGY
(2023)
Article
Multidisciplinary Sciences
Vincent Rouanne, Thomas Costecalde, Alim Louis Benabid, Tetiana Aksenova
Summary: This study demonstrates the feasibility of training and updating BCI decoders during free use of motor BCIs by utilizing an auto-adaptive BCI decoder. By using both the control decoder and MTP decoder, training datasets can be formed in real-time, allowing for real-time adaptation of the BCI. Experimental results show the viability of this auto-adaptive BCI, achieving good results in both discrete and continuous output BCI paradigms.
SCIENTIFIC REPORTS
(2022)
Article
Neurosciences
Carla den Ouden, Andong Zhou, Vinay Mepani, Gyula Kovacs, Rufin Vogels, Daniel Feuerriegel
Summary: Humans and animals can learn and utilize repeating patterns in their environments to form expectations about future sensory events. Predictive coding models have been proposed to explain how learned expectations influence neural activity in the visual system, but there is currently limited evidence for expectation suppression in this system.
Article
Anesthesiology
Juliane Traxler, Andreas von Leupoldt, Johan W. S. Vlaeyen
Summary: Pain can be seen as a signal of bodily error that activates defensive systems and leads to avoidance behavior. However, contrary to findings in anxiety disorders, individuals with higher ERN amplitudes showed lower levels of avoidance behavior during early acquisition and slower learning to avoid pain.
Review
Multidisciplinary Sciences
Julia Gusatovic, Mathias Holsey Gramkow, Steen Gregers Hasselbalch, Kristian Steen Frederiksen
Summary: A systematic review on exercise intervention studies using event-related potentials (ERPs) as outcome for cognitive performance suggests that aerobic exercise interventions have a positive impact on attention, working memory, and inhibition, although the exact neural mechanisms underlying this relationship remain uncertain.
Article
Chemistry, Analytical
Pimwipa Charuthamrong, Pasin Israsena, Solaphat Hemrungrojn, Setha Pan-ngum
Summary: In this study, a visual-ERP-based method was proposed to assess speech discrimination using pictures. Machine learning techniques were employed to classify different task conditions based on features extracted from EEG signals. The results showed that this method achieved high classification accuracy and has the potential to serve as a pre-screening tool to improve the accessibility of speech discrimination assessment.
Article
Computer Science, Hardware & Architecture
Jaeyoung Kang, Behnam Khaleghi, Tajana Rosing, Yeseong Kim
Summary: This paper introduces an efficient GPU-powered framework called OpenHD for mapping HDC applications to GPUs, achieving hardware acceleration. With memory optimization strategies and a novel training method, OpenHD can rapidly achieve the target accuracy and significantly outperform state-of-the-art GPU-powered HDC implementations and non-HDC classification and clustering on GPUs.
IEEE TRANSACTIONS ON COMPUTERS
(2022)
Article
Computer Science, Artificial Intelligence
Gloria Beraldo, Luca Tonin, Jose del R. Millan, Emanuele Menegatti
Summary: This article proposes a novel shared intelligence system for brain-machine interface (BMI) teleoperated mobile robots, where user intention and robot intelligence are both involved in the decision-making process. The experimental results show that the system allows efficient teleoperation of the robot and ensures BMI navigation performances comparable to keyboard control, actively assisting BMI users in accomplishing tasks.
IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS
(2022)
Article
Computer Science, Theory & Methods
Denis Kleyko, Dmitri A. Rachkovskij, Evgeny Osipov, Abbas Rahimi
Summary: This survey provides a comprehensive overview of the computing framework known as Hyperdimensional Computing and Vector Symbolic Architectures (HDC/VSA), including its computational models, applications, and future directions.
ACM COMPUTING SURVEYS
(2023)
Review
Neurosciences
Ji-Hoon Jeong, Jeong-Hyun Cho, Young-Eun Lee, Seo-Hyun Lee, Gi-Hwan Shin, Young-Seok Kweon, Jose del R. Millan, Klaus-Robert Mueller, Seong-Whan Lee
Summary: The brain-computer interface (BCI) is a communication tool that connects the brain with external devices, and has applications beyond communication and control. The 2020 international BCI competition aimed to provide high-quality neuroscientific data for evaluation of BCI advancements. Challenges like few-shot EEG learning, micro-sleep detection, imagined speech decoding, cross-session classification, and ambulatory EEG detection were discussed. The competition attracted participation from scientists and scholars with diverse backgrounds and nationalities, leading to notable BCI advancements and identification of trends for BCI researchers.
FRONTIERS IN HUMAN NEUROSCIENCE
(2022)
Editorial Material
Clinical Neurology
Kristina Simonyan, Stefan K. Ehrlich, Richard Andersen, Jonathan Brumberg, Frank Guenther, Mark Hallett, Matthew A. Howard, Jose Del R. Millan, Richard B. Reilly, Tanja Schultz, Davide Valeriani
MOVEMENT DISORDERS
(2022)
Article
Computer Science, Theory & Methods
Denis Kleyko, Dmitri Rachkovskij, Evgeny Osipov, Abbas Rahimi
Summary: This is Part II of a comprehensive survey on the computing framework known as Hyperdimensional Computing and Vector Symbolic Architectures (HDC/VSA). It provides an overview of existing applications and the role of HDC/VSA in cognitive computing and architectures, as well as directions for future work. The survey aims to be useful for both newcomers and practitioners.
ACM COMPUTING SURVEYS
(2023)
Article
Computer Science, Information Systems
Fabio DellrAgnola, Ping-Keng Jao, Adriana Arza, Ricardo Chavarriaga, Jose del R. Millan, Dario Floreano, David Atienza
Summary: This study proposes a machine learning algorithm for real-time cognitive workload monitoring in drone operations during search and rescue missions. By using multimodal physiological signals and subject-specific optimization, the algorithm can effectively monitor the cognitive workload of rescuers.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Biophysics
Ju-Chun Hsieh, Hussein Alawieh, Yang Li, Fumiaki Iwane, Linran Zhao, Richard Anderson, Syed Ibtisam Abdullah, Kai Wing Kevin Tang, Wenliang Wang, Ilya Pyatnitskiy, Yaoyao Jia, Jose del R. Millan, Huiliang Wang
Summary: Brain-computer interfaces (BCIs) have played a key role in medical applications, but there is a need for more stable and preparati-on-free EEG electrodes. In this study, a long-term stable and low impedance conductive polymer-hydrogel EEG electrode was designed, allowing for long-term and wearable BCIs. The electrode demonstrated good performance in detecting alpha rhythm changes and capturing motor imagery and event-related potentials.
BIOSENSORS & BIOELECTRONICS
(2022)
Article
Engineering, Electrical & Electronic
Robert Guirado, Abbas Rahimi, Geethan Karunaratne, Eduard Alarcon, Abu Sebastian, Sergi Abadal
Summary: Hyperdimensional computing (HDC) is a computing paradigm that uses hypervectors to represent and manipulate data. In-memory computing (IMC) is an efficient hardware platform for executing HDC algorithms. This article introduces WHYPE, a scale-out HDC architecture that uses wireless in-package communication to connect distributed IMC cores, enabling massive parallelization and efficient computation.
IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS
(2023)
Article
Multidisciplinary Sciences
Thibault Porssut, Fumiaki Iwane, Ricardo Chavarriaga, Olaf Blanke, Jose del R. Millan, Ronan Boulic, Bruno Herbelin
Summary: The brain's reaction to a disruption in embodiment, termed Breaks in Embodiment (BiE), was investigated using electroencephalography (EEG). The study found that error-related potentials were observed during the monitoring step when participants experienced a BiE event. Importantly, the EEG signature showed amplified potentials following a non-embodied condition, indicating an accumulation of errors across steps. These neurophysiological findings provide insights into how progressive disruptions impact the expectation of embodiment in a virtual body.
Article
Computer Science, Artificial Intelligence
Michael Hersche, Mustafa Zeqiri, Luca Benini, Abu Sebastian, Abbas Rahimi
Summary: Neither deep neural networks nor symbolic AI has achieved human-level intelligence due to the binding problem in neural networks and the exhaustive rule searches in symbolic AI. Neuro-symbolic AI aims to combine the best of both paradigms, but still faces these problems. However, a proposed neuro-vector-symbolic architecture addresses these issues by utilizing powerful operators on distributed representations. The NVSA achieves new records in accuracy and speed in solving Raven's progressive matrices datasets.
NATURE MACHINE INTELLIGENCE
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Frigyes Samuel Racz, Rawan Fakhreddine, Satyam Kumar, Jose del R. Millan
Summary: In this study, a novel decoding algorithm utilizing Riemannian geometry, template matching and adaptive re-centering was proposed and evaluated for single-trial detection of slow cortical potentials. The results showed that the algorithm can efficiently detect these signals in online applications.
2023 11TH INTERNATIONAL IEEE/EMBS CONFERENCE ON NEURAL ENGINEERING, NER
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Ruofan Liu, Satyam Kumar, Hussein Alawieh, Evan Carnahan, Jose del R. Millan
Summary: To accurately decode motor intentions, motor imagery-based brain-computer interfaces typically require subject-specific calibration data. This paper proposes a geometry-aware deep learning architecture that exploits the spatial similarity of motor imagery neural activity between users. The results show that the proposed method outperforms classical decoding algorithms in a subject-specific setting and achieves similar performance to subject-specific decoders in a transfer learning setting.
2023 11TH INTERNATIONAL IEEE/EMBS CONFERENCE ON NEURAL ENGINEERING, NER
(2023)
Proceedings Paper
Automation & Control Systems
Gloria Beraldo, Luca Tonin, Amedeo Cesta, Emanuele Menegatti, Jose del R. Millan
Summary: Telepresence robots can help people with special needs interact with people and the environment remotely, but the accuracy of alternative communication channels such as brain-machine interfaces is lower. This study compared the navigation performance of a brain-machine interface with a keyboard interface and found similar results, but differences in user inclination were observed in different navigation situations. It suggests the need to adapt the shared intelligence system according to the real-time user's ability and the surrounding environment.
INTELLIGENT AUTONOMOUS SYSTEMS 17, IAS-17
(2023)
Article
Computer Science, Artificial Intelligence
Ping-Keng Jao, Ricardo Chavarriaga, Jose del R. Millan
Summary: The study found that adaptively increasing the difficulty level based on subjective perception through decoding EEG signals is more beneficial than increasing the level at fixed time intervals. To investigate the effectiveness of the EEG decoder, a visuomotor learning task was designed to pilot a simulated drone through waypoints of different sizes. The EEG decoder was compared with a condition where subjects manually regulated the difficulty level. The decoding performance of EEG condition was higher than chance level in 16 out of 26 cases, and the behavioral results were similar to the manual regulation condition.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Denis Kleyko, Geethan Karunaratne, Jan M. Rabaey, Abu Sebastian, Abbas Rahimi
Summary: Memory-augmented neural networks enhance performance by incorporating an external key-value memory, where the complexity is determined by the number of support vectors. We propose a generalized key-value memory that allows for flexible control of the tradeoff between robustness and resource consumption. This is particularly useful for in-memory computing hardware which utilizes nonvolatile memory devices for efficient storage and computation, effectively mitigating nonidealities without the need for neural network retraining.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)