Abstract: In order to study EEG for fatigue driving, this article collects and extracts data from experiments using wavelet transform. This article is based on the subjects signs and brain electrical device collection pretreatments. The mean amplitudes of four wave bands of α wave, β wave, θ wave, and δ wave, and other indicators (α+β)/β,α/β,(δ+α)/(α+β),(α+β)/θ,  are extracted by the wavelet transform experiment, total eight synthetic indicators are integrated into EEG characteristic parameters. The experiment extracts principal component feature information with a contribution rate of 90% or more by KPCA to form characteristic set, and the feature information is input into least square support vector machine(LSSVM). The KPCA-LSSVM prediction model is established and compared with other four model tests. Finally, the model obtains the average correct rate of 89.47%. The comparative experiment proves the effectiveness of the experiment and its advantages in data processing speed.

Key words: wavelet transform, kernel principal component analysis, least square support vector machine, electro encephalo gram