Nowadays, the world needs safe and smart machines that can prevent human errors in different situations. Stress is an important factor in accidents which causes the human error. Many accidents can be prevented by identifying the stress of the driver and warning them. Due to its complexity, identifying stress in drivers is only possible by intelligent algorithms. In this paper, the Electrocardiogram (ECG) signal from drivedb dataset is used to detect stress in drivers, which has useful information that can be recorded more easily while driving. Afterwards, with a set of statistical, entropy, morphology, and chaos features, useful information is extracted from the signal. Then, in order to optimize the features, the Relief feature selector is used. Optimal features information is evaluated using Artificial Neural Networks (ANNs). Using the proposed method, the stress in drivers is detected with an accuracy of 92.6%, which has increased classification accuracy compared to recent researches. Manuscript profile