Topological feature extraction of nonlinear signals and trajectories and its application in EEG signals classification

This study introduces seven topological features that characterize attractor dynamic of nonlinear and chaotic trajectories in a phase space. These features quantify volume, occupied space, nonuniformity, and curvature of trajectory. The features are evaluated as initial point invariant measures by a practical approach, which means that a feature is only sensitive to dynamic changes. The Lorenz and Rossler system trajectories are employed in this evaluation. Moreover, the proposed features are used in a real world application, i.e. epileptic seizure electroencephalogram signal classification. As the result shows, these features are efficient in this task in comparison with others studies that used the same dataset and evaluation method.