K-Complexes Detection in Sleep Electroencephalogram Signals Utilising Fast Fourier Transform Based Second Order Difference Plot
The intention of this paper was to describe the design, improvement and verification of the performance of the K-complexes detection system in Electroencephalogram (EEG) signals. This work proposed a mechanism for features extraction based on merge the Fast Fourier Transform (FFT) with second-order deference plots (SODPs), for solving many-objective optimization problems for the high dimensionality of any database. In the first phase of the model design, an EEG signal was divided into relatively small intervals or segments. FFT was implemented for each EEG segment. To find out the most effective input features to represent the EEG signal, the SODPs were utilized. The extracted features were then utilised as the input to various classifiers, such as K-means, the Naïve Bayes algorithm and least square support vector machines (LS-SVM). The results were compared with existing studies, the proposed approach supplies a high rate of accuracy, ~95.35%. As conclusions outcomes showed that the proposed approach can develop the classification of KCs in EEG signals. The presented method provided the best outcomes compared with others. In addition, the proposed method can have practical implications in assisting physicians to detect transient events in sleep stages more accurately than the existing methods. The new approach can be applied to various medical data types, for example, restless legs syndrome, epilepsy and obstructive sleep apnea, etc.
Electroencephalogram. K-complexes. Fast Fourier Transform. Second Order Deference Plots. Segmentation Technique.