Classification of Electroencephalography Signals with Auditory Stimuli using Complex Network

Electroencephalography signals serve as valuable tools for studying the human brain activity and detecting various neurological disorders. This study proposes a novel approach to analyze electroencephalography signals in response to auditory stimuli and language variations utilizing Visibility Graph and Horizontal Visibility Graph algorithms, with and without segmentation. By extracting statistical properties and fitting the degree distribution of the networks, a feature vector is obtained for classification problems using k-nearest neighbor and support vector machine algorithms. Analysis results indicated that the Horizontal Visibility Graph outperformed the VG, whereas segmentation methods showing minimal impact on the classification performance. Moreover, the support vector machine classifiers demonstrate stable performance, and achieve average accuracy of nearly 60% to over 70% through 10-fold cross-validation. This proposed method is the foundation for further studies of pattern classification with EEG signals.