The Use of SPOD and Spherical Harmonics for the Analysis of EEG Data.

The assessment of mental workload from electroencephalogram (EEG) data for brain-computer interfaces (BCI) poses some challenges due to the interaction of spatial and temporal features within the data. Similar challenges are well known in the analysis of turbulent flows, which exhibit complex space-time correlations resulting from large-scale structures. The similarities motivated us to conduct this feasibility work of applying analytic methods of fluid dynamics – i.e., a scheme combining spectral proper orthogonal decomposition (SPOD) and spherical harmonics as basis functions – to EEG data for identifying features representing mental states. Based on EEG data of an existing BCI Hackathon, the scheme yielded some relevant features across subjects and sessions by relying only on a Fourier transform in time and the basis functions in space. The features were then classified by employing a conventional support vector machine algorithm to produce an accuracy comparable to those reported in a previous study on the same EEG data. This performance comparability indicates the scheme's potential for analyzing EEG data in BCI applications. Nevertheless, future work is needed to select specific features as general indicators of mental workload.