Cognitive load estimation on the basis of video information using recurrent neural networks

In this work, the method of estimating cognitive workload is proposed. It is based on the idea to estimate workload using the video information from a camera with recurrent neural networks trained individually. To build a model, the workload is preliminarily estimated under special experimental conditions using task-based approaches while facial and gaze features are extracted from the video during the experiment. Using extracted information and workload estimation as training data cognitive workload is then modeled with recurrent neural networks with long short-term memory.