Fast Fourier Transform Coupled with Machine Learning Algorithm for K-Complexes Detection

This paper proposes a novel K-complexes (KCs) detection approach using sleep electroencephalogram (EEG) recordings. A segmentation technique is used to partition an EEG signal into intervals. Then, fast Fourier transform (FFT) is applied to each EEG segment. To find out the most effective input features to represent the EEG signal, the FFT coefficients were investigated. The extracted features are then utilized as the input to an ensemble classifier which is designed using three classifiers: K-means, the Naive Bayes algorithm and least square support vector machines (LS-SVM). A comparison with existing studies is made and the results showed that the proposed model outperformed state of the art. The proposed approach can be developed as an online system to detect KCs in EEG signals. In addition, it can be applied to other EEG data such as detect sleep apnoea.