Recurrence Analysis of EEG Power and HRV Time Series for Asynchronous BCI Control

Abstract

Autonomic control evidenced in both heart rate variability (HRV) and electroencephalographic (EEG) oscillations, reflect the behaviour of the underlying physiological non-stationary dynamical systems. In order to assess the influence of autonomic changes in the performance of an asynchronous brain-computer interface, recurrence analysis of EEG spectral density time series and HRV feature sets were used in binary classifiers to detect rest state from mental calculation state. Results suggest that recurrence indices of HRV might contribute to improve activity episodes detection for BCI control as the highest performance was achieved with power spectral density features of EEG in combination with recurrence HRV features (AUROC $$=0.81\pm 0.07$$ ).