A multi-modal feature fusion method for pilot expression recognition

Facial expression recognition technology possesses important utility worth in pilot emotion and flight safety monitoring. In order to resolve the issue of insufficient information and inferior recognition accuracy of single modal features in facial expression recognition, a multi-modal feature fusion method for pilot expression recognition is proposed. Laboratory findings show that the recognition precision of the presented method on CK+ and our newly-proposed pilot expression dataset reaches 99.68% and 98.38% respectively, which can not only effectively identify pilot expression, but also outperform other competing methods.