EEG Analysis Using HHT: One Step Toward Automatic Drowsiness Scoring

This paper proposes an algorithm for automatic location of alpha and theta waves in electroencephalogram. This algorithm is a part of developments that aim to process EEG and electroocculogram in order to estimate the drowsiness level of active subjects.. Our algorithm is based on a method recently developed to analyse non-stationary signals: Hilbert Huang transform (HHT). This transform proposes to decompose multi-modal signals into a sum of mono- contribution functions called intrinsic mode functions, then to use the Hilbert transform to compute the instantaneous frequency of each IMF. After a brief review of HHT principles, we propose a qualitative analysis of Hilbert transform accuracy and a method to decrease computation errors that appears when amplitude of the analysed signal is small. The last section of this paper presents the algorithm proposed to locate alpha and theta waves and preliminary results.