Digital Signal Processing and System Theory

Talk Lars Lindemann


lali

Real-time Detection of Event-Related Potentials by Means of EEG

Date: 24.07.2014, 13:00 h - 13:30 h
Room: Aquarium

Lars Lindemann
CAU Kiel, Germany




Details



Current researches at the Christian-Albrechts-University of Kiel have been focussed on artifact based feature extraction. In connection with a Brain-Computer-Interface, these features have been used. An enhancement of this Brain-Computer-Interface is not part of this work. First of all, a 24 channel electroencephalogram of the company Demetec has been connected with the Kiel Real-time Audio Toolkit, which is used for real-time processing. This first step includes a samplerate conversion and implementation of digital filters for noise reduction. In favour of thought extraction, there have been several MATLAB investigations of Body Move- ments and Visual-Evoked Potentials. During this work, the awareness of signal averaging, signal- to-noise level and artifacts has been awaked. The need for low noise disturbance and training before real-time usage is a result of this analysis. Based on this discoveries, a visual-evoked P300 feature classificator has been implemented. In a foregoing training, a training vector is calculated by applying Fisher’s Linear Discriminant Analysis. In real-time processing, the resulting training vector is linear combined with a feature vector for the detection of P300 waves, which are released due to a visual stimuli. The implemented system has an accuracy of 77.8 % with a detection time of about 40 seconds per feature.