Abstarct: Fatigue driving is one of the high-frequency factors in traffic accidents all over the world. Therefore, it is necessary to design ways to identify fatigue for drivers. The signal processing method has helped complete different aspects of the problem and is recognised as a model of brain signals. Previous studies usually use a fast fourier transform (FFT) or discrete wavelet transform (DWT) as a feature selection technique. In this thesis, a novel method for detection and elimination of eyeblxink artifacts and drowsiness detection of the driver from the electrodes of the frontal region has been proposed. This method baxsed on the ensemble empirical mode decomposition (EEMD) and hilbert transform on the recorded signals emitted by the EEG. These EEG signals are decomposed to extract the intrinsic mode functions of the (IMFs) using the EEMD technique. Then the IMF components are determined using Hilbert transform and characterized by four classification criteria in multiple tests. in this thesis, we use an existing databaxse containing 34 EEG collection electrodes. The five - minute recording of the subject was considered as normal EEG data. In simulation, we have improved a method baxsed on EMD and four - scale entropy. We increased the accuracy of 88/74% to 92/86% and obtained sensitivity of 90/61% and specificity 94/15% by removing the eyeblxink artifacts by adaptive filter and replacing the EEMD and extracting features by hilbert transform.