Research on a Fatigue Detection Method Based on Phoneme

In order to effectively detect the fatigue status of train drivers and ensure the safety of train operation, in view of the limitations of existing methods in feature selection and model construction, a driving fatigue detection method based on phoneme spectrogram and convolutional neural network was proposed. According to the characteristics of the work tasks of the train driver, a fatigue speech database was constructed and the effectiveness of the database was verified; phonemes that can express fatigue information in a more detailed and intuitive manner were selected as the research object, and based on the speech signal processing theory and the deep learning theory, a driving fatigue detection model was established to achieve a more accurate and robust driving fatigue detection. The experimental results show that the precision, recall and accuracy (96.7%) of this method are better than the existing methods. The research on the train driver fatigue state automatic detection technology carried out in this paper not only has a wide application prospect in the intelligent rail transit industry, but also can provide theoretical and technical support for the research on human-machine adaptive interaction of trains.