Delaunay Triangulated Simplicial Complex Generation for EEG Signal Classification
Srikireddy Dhanunjay Reddy, Tharun Kumar Reddy · IEEE Sensors Letters · 2024
This letter proposes the novel delaunay triangulated simplicial complex generation (DTSG) framework for electroencephalogram (EEG) sensors signal classification with the help of persistent homological features. Using the DTSG framework, EEG sensor signals are mapped into the point clouds by implementing the multidimensional scaling approach. Further, simplicial complexes are generated using the Delaunay triangulation method for the extraction of persistent homological features. These homological features help in finding the underlying complex connectivity and structural behavior of nonstationary overlapping neurological disorders, such as depression, bipolar disorder, and schizophrenia. The proposed method is implemented and evaluated on a publicly available major depressive disorder dataset to classify depressive patients from healthy controls. The proposed DTSG framework has shown its efficacy in both binary and multiclass depression classification tasks with 96.54% and 85.67% accuracies, in comparison with existing methods by utilizing the extracted persistent homological features.