Prediction Method For The Detection of Epilepsy In The EEG Signals

The distinctiveness of EEG signals, which are a component of human biological signals, enables us to detect abnormal neurological activity, such as epilepsy, through pre-processing. Over the years, the use of EEG signal processing for predicting and detecting epilepsy has received significant attention. A straightforward method for recognizing epileptic events is proposed based on the recorded EEG signals. The linear prediction filter is applied to monitor seizure activity and predict the occurrence of spikes and sharp waves. This study analyzes the linear prediction mechanism by determining coefficients for each set of windows, with the goal of finding the superior model. The performance of the model is evaluated using the error value of the predicted signal. The results show that the energy of the prediction error signal during seizures is higher than during seizure-free periods, allowing for the identification and location of seizure intervals using the energy value. The proposed approach is evaluated using Receiver Operating Characteristics (ROC) analysis and outperforms existing methods.