Detecting driving drowsiness using EEG, EOG and driving quality signals

Present study has sought to investigate the drowsiness so that a reduction in the casualties due to the road accidents can be achieved. In order to reduce the eye blink artifact, an automatic mechanism based on the ICA method and Higuchi’s fractal dimension has been applied. After feature extraction, for selecting the best subset of features, a new combined method, called CSFS-SFS, has been developed. This method reduces the time of calculations up to 69.21% and keeps the classification results almost unchanged. For diagnosis of the drowsiness and evaluation of the data labeling criteria, a new idea based on SOM network has been used. SOM network has been trained and tested using the obtained data from two classes of AS (awareness state) and DS (drowsiness state). The accuracy of 76.51±3.43% verifies the reliability of the considered criteria for AS and DS data labeling. By the implementation of the suggested idea, the need for expert labeling is significantly removed.