Developing new techniques to analyse and classify EEG signals

A massive amount of biomedical time series data such as Electroencephalograph (EEG), electrocardiography (ECG), Electromyography (EMG) signals are recorded daily to monitor human performance and diagnose different brain diseases. Effectively and accurately analysing these biomedical records is considered a challenge for researchers. Developing new techniques to analyse and classify these signals can help manage, inspect and diagnose these signals. In this thesis novel methods are proposed for EEG signals classification and analysis based on complex networks, a statistical model and spectral graph wavelet transform. Different complex networks attributes were employed and studied in this thesis to investigate the main relationship between behaviours of EEG signals and changes in networks attributes. Three types of EEG signals were investigated and analysed; sleep stages, epileptic and anaesthesia. The obtained results demonstrated the effectiveness of the proposed methods for analysing these three EEG signals types. The methods developed were applied to score sleep stages EEG signals, and to analyse epileptic, as well as anaesthesia EEG signals. The outcomes of the project will help support experts in the relevant medical fields and decrease the cost of diagnosing brain diseases