Showing papers by "Gernot Müller-Putz published in 2011"

Tools for Brain-Computer Interaction: A General Concept for a Hybrid BCI.

Abstract: The aim of this work is to present the development of a hybrid Brain-Computer Interface (hBCI) which combines existing input devices with a BCI. Thereby, the BCI should be available if the user wishes to extend the types of inputs available to an assistive technology system, but the user can also choose not to use the BCI at all; the BCI is active in the background. The hBCI might decide on the one hand which input channel(s) offer the most reliable signal(s) and switch between input channels to improve information transfer rate, usability, or other factors, or on the other hand fuse various input channels. One major goal therefore is to bring the BCI technology to a level where it can be used in a maximum number of scenarios in a simple way. To achieve this, it is of great importance that the hBCI is able to operate reliably for long periods, recognizing and adapting to changes as it does so. This goal is only possible if many different subsystems in the hBCI can work together. Since one research institute alone cannot provide such different functionality, collaboration between institutes is necessary. To allow for such a collaboration, a new concept and common software framework is introduced. It consists of four interfaces connecting the classical BCI modules: signal acquisition, preprocessing, feature extraction, classification, and the application. But it provides also the concept of fusion and shared control. In a proof of concept, the functionality of the proposed system was demonstrated.

Combined motor imagery and SSVEP based BCI control of a 2 DoF artificial upper limb

Abstract: A Brain–Computer Interface (BCI) is a device that transforms brain signals, which are intentionally modulated by a user, into control commands. BCIs based on motor imagery (MI) and steady-state visual evoked potentials (SSVEP) can partially restore motor control in spinal cord injured patients. To determine whether these BCIs can be combined for grasp and elbow function control independently, we investigated a control method where the beta rebound after brisk feet MI is used to control the grasp function, and a two-class SSVEP-BCI the elbow function of a 2 degrees-of-freedom artificial upper limb. Subjective preferences for the BCI control were assessed with a questionnaire. The results of the initial evaluation of the system suggests that this is feasible.

First Steps Toward a Motor Imagery Based Stroke BCI: New Strategy to Set up a Classifier

Abstract: A new approach in motor rehabilitation after stroke is to use motor imagery (MI). To give feedback on MI performance BCIs can be used. This requires a fast and easy acquisition of a reliable classifier. Usually, for training a classifier, EEG data of motor imagery without feedback is used, but it would be advantageous if we could give feedback right from the beginning. The sensorimotor EEG changes of the motor cortex during active and passive movement and motor imagery are similar. The aim of this study is to explore, whether it is possible to use EEG data from active or passive movement to set up a classifier for the detection of motor imagery in a group of elderly persons. In addition, the activation patterns of the motor cortical areas of elderly persons were analysed during different motor tasks. EEG was recorded from three Laplacian channels over the sensorimotor cortex in a sample of 19 healthy elderly volunteers. Participants performed three different tasks in consecutive order, passive, active hand movement and hand motor imagery. Classifiers were calculated with data of every task. These classifiers were then used to detect ERD in the motor imagery data. ERD values, related to the different tasks, were calculated and analysed statistically. The performance of classifiers calculated from passive and active hand movement data did not differ significantly regarding the classification accuracy for detecting motor imagery. The EEG patterns of the motor cortical areas during the different tasks was similar to the patterns normally found in younger persons but more widespread regarding localization and frequency range of the ERD. In this study, we have shown that it is possible to use classifiers calculated with data from passive and active hand movement to detect motor imagery. Hence, for working with stroke patients, a physiotherapy session could be used to obtain data for classifier setup and the BCI rehabilitation training could start immediately.

Motivation Influences Performance in SMR-BCI

Abstract: In this study we investigated the effect of motivation on performance when using a BrainComputer Interface (BCI) based on sensorimotor rhythms (SMR). After pooling the data acquired with four different SMR-BCI protocols in one sample of N=41 participants, we found a positive correlation between the motivational components ”challenge” and ”incompetence fear” and accuracy in percent correct responses. As motivation seems to have a positive effect on SMR-BCI performance, we recommend to enhance motivation if possible and to monitor motivation in BCI settings.

Design of a novel covert SSVEP-based BCI

Abstract: Brain computer interfaces (BCI) employing steady-state visually evoked potential (SSVEP) modulations have been investigated increasingly in the last years because of their high signalto-noise ratio and information transfer rate. However, independent SSVEP BCI based on covert attention show a drop in robustness which makes it difficult to use on patients with impaired or nonexistent ocular motor control. In the present paper, offline analysis is aimed at investigating the influence of three important parameters on the performance of covert SSVEP BCI : feature extraction algorithms, window length and number of harmonics. We also proposed a new ”checkerboard” pattern and compared its performance with lines pattern. We have shown that the use of this pattern and only one harmonic yielded an average accuracy of approximately 79% across five subjects (with four subjects at more than 81%) with 6s-window length and feature extraction algorithm based on canonical correlation analysis or lock-in analyzer system. The short 5 or 6s-concentration time, the absence of training due to the use of only one harmonic, the robustness make this method very well suited for detecting command following and testing communication in unresponsive post-comatose patients.

Somatosensory evoked potentials elicited by stimulating two fingers from one hand — Usable for BCI?

Abstract: Steady-state somatosensory evoked potentials (SSSEPs) have been elicited using vibro-tactile stimulation on two fingers of the right hand. Fourteen healthy subjects participated in this study. A screening session, stimulating each participant's thumb, was conducted to determine individual optimal resonance-like frequencies. After this screening session, two stimulation frequencies per subject were selected. Stimulation was then applied simultaneously on the participant's thumbs and middle finger. It was investigated whether it is possible to classify SSSEP changes based on an attention modulation task to determine possible BCI applications. A cue indicated the participants to shift their attention to either the thumb or the middle finger. Offline classification with a lock-in analyzer system (LAS) and a linear discriminant analysis (LDA) classifier was performed. One bipolar channel and no further optimization methods were used. All participants except one reached classification results above chance level classifying a reference period without focused attention against focused attention either to the thumb or the middle finger. Only two subjects reached accuracies above chance, classifying focused attention to the thumb vs. attention to the middle finger.

Comparison of Feature Extraction Methods for Brain-Computer Interfaces

Abstract: Classification accuracies of feature extraction methods (FEMs) as used in sensory motor rhythm (SMR) based Brain-Computer Interfaces (BCIs) were compared offline. Features were extracted from 9 subjects and classified with linear discriminant analysis (LDA). The following FEMs were compared: adaptive autoregressive parameters (AAR), bilinear AAR (BAAR), multivariate AAR (MVAAR), band power (BP), phase locking value (PLV), time domain parameters (TDP), and Hjorth parameters. Most FEMs contain meta parameters and it is crucial to tune these meta parameters carefully to tap the full potential of these methods. Therefore, all meta parameters were optimized in a subject-specific way with a genetic algorithm (GA) [1].

A concept to standardize raw biosignal transmission for brain-computer interfaces

Abstract: With this concept we introduced the attempt of a standardized interface called TiA to transmit raw biosignals. TiA is able to deal with multirate and block-oriented data transmission. Data is distinguished by different signal types (e.g., EEG, EOG, NIRS, …), whereby those signals can be acquired at the same time from different acquisition devices. TiA is built as a client-server model. Multiple clients can connect to one server. Information is exchanged via a control- and a separated data connection. Control commands and meta information are transmitted over the control connection. Raw biosignal data is delivered using the data connection in a unidirectional way. For this purpose a standardized handshaking protocol and raw data packet have been developed. Thus, an abstraction layer between hardware devices and data processing was evolved facilitating standardization.

Brisk movement imagination for the non-invasive control of neuroprostheses: A first attempt

Abstract: The consequences of a spinal cord injury (SCI) are tremendous for the patients. The loss of motor functions, especially of grasping, leads to a dramatic decrease in quality of life. With the help of neuroprostheses, the grasp function can be substantially improved in cervical SCI patients. Nowadays, systems for grasp restoration can only be used by patients with preserved voluntary shoulder and elbow function. In patients with lesions above the 5th vertebra, not only the voluntary movements of the elbow are restricted, but also the overall number of preserved movements available for control purposes decreases. In this work, a new method for the non-invasive use of a Brain-Computer Interface (BCI) for the control of the hand and elbow function is presented.

TOBI Interface A (TiA) – A Standardized Interface to Transmit Raw Biosignals

Abstract: TOBI interface A (TiA) describes a standardized interface to transmit raw biosignals, supporting multirate and block-oriented transmission of different kinds of signals at the same time. To facilitate a distinction between those kinds of signals, so-called signal types are introduced. TiA is a single server, multiple client system, whereby clients can connect to the server at runtime. Meta information transfer between client and server is divided into a control- and data connection. The control communication is using TCP with XML messages, and data transmission is using UDP or TCP with a binary data stream.

TiA – Standardizing Raw Biosignal Delivery in BCIs

Abstract: TiA is a concept to transmit raw bisoignals in a standardized way for BCI purposes. It provides the possibility for multirate and and block-oriented data transmission. Different kinds of signals, divided into so called signal types (e.g., EEG, EMG, ECG) can be transmitted at the same time, whereby a data distinction is always possible. TiA utilizes a client–server principle with one server performing data acquisition and multiple clients as data consumers. Data is divided into immutable meta information and raw biosignals. A standardized handshaking protocol and TiA data packets were introduced. Applying this concept an abstraction layer is evolved facilitating the standardization process for BCI development.