Application of genetic algorithm based support vector machine in selection of new EEG rhythms for drowsiness detection

•GA-SVM is used to find the optimal rhythms for drowsiness detection.•The original EEG signals are decomposed using wavelet packet transform.•The drowsiness detection effect of the new rhythm was evaluated using LOSO-CV. The electroencephalogram (EEG) signals are important for drowsiness detection. However, in some specific application scenarios, whether there is a more accurate rhythm for drowsiness detection is worth further study. Therefore, a method of finding the optimal EEG rhythm for drowsiness detection using the genetic algorithm based support vector machine (GA-SVM) has been proposed in this study. This study used the original EEG signals in the Sleep EDF [Expanded] database for analysis and experiments. First, the original signals were divided into several epochs, and the signals of each epoch were decomposed using db10 wavelet packet transform and haar wavelet packet transform, respectively. Then, the GA-SVM was used to select the most accurate rhythm for drowsiness detection. Finally, leave-one-subject-out cross-validation (LOSO-CV) was used to evaluate the performance of each rhythm for drowsiness detection. The results show that the gamma rhythm has the best detection efficiency in the five traditional rhythms, and the accuracy rate is 80.94%. The detection accuracy of the new rhythm Rhythm (III) (43.75–48.046875 Hz) proposed in this study is 89.52%. The new rhythm proposed in this study showed bast performance in drowsiness detection.