UTDrive: Emotion and Cognitive Load Classification for In-Vehicle Scenarios

Emotions and non-driving related cognitive tasks affect a driver’s control over a vehicle and may result in driving errors and traffic accidents. Presence of a monitoring device that would assess driver’s state could help reduce such errors by providing the driver with alerts and directing other in-vehicle active safety devices. The focus of this study is on the evaluation of speech production-based and cepstral-based acoustic features for the task of emotion and cognitive load classification in real driving scenarios. The newly proposed classifiers utilize support vector machine (SVM) based fusion of raw features and Gaussian mixture model (GMM) scores and provide classification performance of 79 % and 95.2 % in the task of neutral vs. negative emotion classification and two cognitive tasks classification, respectively.