Combining Principal Component Analysis and Template Matching for Effective EEG Signal Classification in Seizure Detection

EEG is routinely used in the assessment of the neurological patient’s state, especially in identifying the episodes of seizure in cases of epilepsy. However, the classification of the training signals into item classes is not always accurate because of the fact that they are based on EEG signals and the data is high dimensionality and complexity. This research work develops an original algorithm incorporating the PCA and template matching models for accurate EEG indicator organization in seizure recognition. PCA is used to transform the EEG signals into a space where relevant features are retained, and noise is minimized. Then, the signals are classified by using template matching technique with seizure and non-seizure templates. The integration of these two methods improves the detection performance and reduces the computational time when analysing a large amount of seizure data as shown from the experimental outcomes using benchmark EEG data. The approach proposed here has potential for real time seizure monitoring and can provide potential avenue towards enhanced clinical diagnosis and patient care.