Wavelet-based feature extraction for EEG classification

This paper uses the wavelet transform to obtain descriptive features of EEG data from a prenatal drug exposure study. The interest is in analyzing the data to determine temporal and spectral information that can be used as a predictor of central nervous system dysfunction. While the FFT can be useful, it does not incorporate temporal information. The wavelet transform is applied because it provides a multiresolution decomposition in time and frequency and is therefore well suited for the analysis of transient nonstationary signals such as these. Results show that useful features are obtained from the histograms of the relative time durations of specific frequencies obtained via the wavelet transform