Beta 1


Title Towards 24-7 Brain Mapping Technology
Author Nielsen, Brian
Supervisor Hansen, Lars Kai (Intelligent Signal Processing, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2009
Abstract The use of closely spaced subcutaneously implanted electrodes for EEG recording is examined. A comparison between conventional electrodes and subcutaneous electrodes is made. Only a limited amount of data material is available. Several methods are employed for the comparison: frequency spectra, ERP, the amount of artifacts, and ICA decomposition. The analysis shows that the data recorded from the two different recording methods is almost identical, although some differences are found. The found differences do not give a clear picture of whether the subcutaneous electrodes provide better or worse data compared to the conventional electrodes. A classification of two different data sets is done in order to investigate the use of a limited amount of electrodes: 1) Classification of a data set containing visual evoked potential (VEP) trials is performed by three different classification methods: Fisher’s linear discriminant (FLD), linear support vector machines (SVM), and Gaussian SVM. A good classification from the supplied data is not possible. 2) Classification of a data set containing tasks based on motor imagery is performed. FLD, linear SVM, and Gaussian SVM are used as classifiers. Feature extraction is performed on the basis of event related potential (ERP) and event related spectral perturbation (ERSP). Using only four electrodes a classification accuracy of 94% is obtained. The results from the second classification show that it is possible to perform a successful classification using only a few electrodes.
Series IMM-M.Sc.-2009-17
Fulltext
Original PDF ep09_17.pdf (3.12 MB)
Admin Creation date: 2009-03-23    Update date: 2010-10-28    Source: dtu    ID: 240392    Original MXD