Driver drowsiness is recognized as one of the leading causes of accidents and road incidents globally. In recent years, considerable efforts have been directed towards developing effective tools for detecting and predicting driver drowsiness using cost-effective methods suitable for public use. Biological signals, particularly electroencephalogram (EEG), have become highly valued for their immediate reflection of drowsiness-induced changes in detecting driver alertness and fatigue. Despite significant advancements in drowsiness detection, researchers continue to strive for enhanced accuracy in detection models by extracting novel and relevant features from EEG signals. In this study, we introduce a new set of features derived from EEG signals and develop an algorithm to predict driver drowsiness using these features. The dataset used in this study was gathered from 50 volunteers during driving activity in the Nasir driving simulator, employing consumer-grade headsets Muse 2 and Muse S to record EEG signals. Following preprocessing and segmentation of the signals into 30-second epochs, features including the zero-crossing rate of the original signal, as well as its first and second derivatives, were extracted. Statistical analysis identified 11 features which exhibited significant differences between the states of alertness and drowsiness. These features were then utilized to develop a predictive model employing KNN and SVM classifiers. The maximum accuracy achieved in predicting driver drowsiness using the proposed algorithm was 86.37%. Hence, the introduced features are proposed as effective parameters for enhancing the accuracy of drowsiness prediction.