AG Brain-Machine Interface

BCI demonstration at Open Lab event

In May, a delegation from the Department of Psychology, Justus-Liebig-University Giessen, visited the LIN to further push collaborative work already started last year. The scientific highlight was the symposium "Resources of perception, action and learning" which was very inspiring to all attendies. The BMI group exhibited a binary brain-computer interface (BCI) that enables people to respond to questions only by brain activity induced by attention. All out of 18 questions were correctly decoded by the BCI, indicating that it worked perfect with the participant while average accuracy in previous experiments was about 90%.

Honorary symposium

On Feb. 22, 2023, an honorary symposium was held on the occasion of the retirement of Hermann Hinrichs, who was head of the AG BMI since its inception. Among others, companions such as Thomas Münte (University of Lübeck), Ariel Schönfeld (Kliniken Schmieder Heidelberg) and Steven Hillyard (UC San Diego) presented milestones of their joint research work and told one or the other anecdote. George Mangun from UC Davis was welcomed as a surprise speaker. One of the highlights was the speech held by Hans-Jochen Heinze, head of the department, and was undoubtedly very entertaining and emotional for all involved.

Long Night of Science

After the Long Night of Science 2021 only took place online, interested people could finally visit the research facilities in person again in 2022. The AG BMI participated with a combination of lecture and demonstration of a Brain-Computer Interface (BCI). The BCI allowed the volunteer participant to answer “yes” and “no” to spontaneously asked questions. To do so, she only had to pay attention to an the color stimulus linked to the response (green=”yes”, red=”no”), without any muscle involvement. The answer recognized by the BCI from the electrical brain activity was fed back via loudspeaker. This impressively demonstrated to the audience how reliably communication can be realized solely by paying attention to a visual stimulus.

Since its foundation, the working group has been headed by Prof. Dr.-Ing. Hermann Hinrichs, who retired in February 2023. Since March 2023, the AG is headed by Dr. Christoph Reichert.

Christoph Reichert studied computer science at the Otto von Guericke University Magdeburg from 2001 to 2007. Afterwards, he worked as a research assistant at the University Clinic for Neurology in Magdeburg. He has been employed at LIN since 2015. He completed his PhD in 2016 at the Faculty of Computer Science at Otto-von-Guericke University.

His work at the LIN was accompanied by a close cooperation with the STIMULATE research campus, where he led the Brain-Machine Interfaces research group from 2018 to 2019. Since 2022, he is one of the postdoc representatives at LIN. His main interest is in the application of machine learning for decoding electrophysiological brain signals.

Visual spatial attention as a control signal

In this project we develop novel paradigms that permit to generate control signals, simply by directing attention to peripherally presented symbols (N2pc paradigm), independent of eye movemements. To achieve this, we decode the electroencephalogram using an efficient dedicated algorithm. In the long run, we intend to probe this procedure to provide an instrument of communication for severely paralyzed patients.

Unidirectional Communication using collaborative BMIs

Collaborative BMIs decode brain activity of different persons who pursue the same goal at the same time, i.e. they try to generate the same control signal using their brain activity. In this project we aim at extending this idea and utilize the synchrony and asynchrony of two simultaneously recorded electroencephalograms to transmit information from one person to another person. Our goal is to show that covert communication, i.e. unrecognizable to a third person, is possible and that this kind of communication is more reliable than common approaches in communication BMIs.

Completed Projects

Diagnostic Glove: Disease Diagnoses in daily life from wearable kinematics

A cooperation with the University Hospital’s Department of Neurology and the IKND

The aim of this project was to use data gloves (gloves or exoskeletons equipped with sensors) to detect specific motor diseases in early stages. Some motor diseases can subjectively not be differentiated in early stages but their prognoses are differently severe. Using sensor data in combination with machine learning techniques, we identified typical patterns associated with natural hand movements at different ages. The dynamically variable nature of individual movements was particularly challenging for decoding. Finally, we examined hand movement data from ALS patients and a control group to characterize typical patterns of impairment specific to this disease.

Brain-Machine Interface – OP planning and selection procedures (as subproject of the research campus STIMULATE)

The project is concerned with the investigation of paradigms and algorithms that make it possible to identify brain function for robust BMI control in different forms of brain lesion after stroke. On the one hand, this includes the individual optimal localization of non-invasive or invasive electrodes (grids) and, on the other hand, enables the detection of intended actions in the single triangle, i.e. in real-time, and the generation of machine commands. Therefore, in this work package, selection procedures are to be developed which allow a simple intuitive control of an autonomously acting robot arm.

In-Ear-BMI (as subproject of the research campus STIMULATE)

This project comprises the development and implementation of hardware and software components for a miniaturized system for the acquisition of EEG signals from the auditory canal, realized as a smartphone extension. Research goals include the development of suitable electrodes and electrode localizations, dedicated miniaturized amplifiers/signal converters and smartphone-based analysis software that can extract specific brain function parameters from the EEG based on neuroscientific approaches.