| Summary: | In this thesis, we consider three procedures to extract the mismatch negativity, a component of event-related potential, from electroencephalography data: optimal digital filtering, wavelet decomposition, and independent component analysis decomposition procedures. The procedures are compared on two different datasets, stressing their advantages over the conventional difference wave procedure. The main results of the thesis support the use of the wavelet decomposition and independent component analysis decomposition procedures to reveal the experimental effects which are expected from the literature, but not distinguishable through the traditional procedure, and show that these developed procedures may allow us to reduce the duration of an experimental session. Also, we discuss some practical issues related to the use of independent component analysis-based procedures in the extraction of the mismatch negativity. Finally, we consider a method for spatial denoising in multi-channel electroencephalography data, which can be used as a preprocessing step prior to the extraction of the mismatch negativity or any event-related potential as well.
|
|---|