A Pilot Multimodal Physiological Signals Dataset for Negative Emotion Recognition in Simulated Flight Task Scenarios

Pilots' emotional fluctuations, which may impair their cognitive processes and increase risk-taking behaviors during flight operations, are one of the important human factors affecting aviation safety. Thus, it is a vital step to recognize different emotional states, especially a few negative emotions such as stress or panic, in real time for decreasing the risk of aviation accidents. However, there are no open and high-confident datasets about pilots' emotional states due to the difficulty of data acquisition, which made it a challenge for related studies of emotion recognition. In this paper, we collected 70,000 samples of multimodal physiological signals and QAR data from flight simulator for establishing a Pilot Multimodal Physiological Signals (Pilot-Multi-PS) dataset. Each of 17 cadet pilots operated a high-fidelity flight simulator in both normal and emergency flight maneuver scenarios, eliciting three emotions: stress, panic, and calmness. We believe that Pilot-Multi-PS represents a valuable resource for human factors analysis, facilitating the improvement of pilots' training performance.