Hidden Markov Models to identify pilot instrument scanning and attention patterns

A previous study successfully demonstrated the potential usefulness of Hidden Markov Model (HMM) analysis techniques in analyzing pilots' eye-movement data to detect differences in scanning and attention patterns caused by display format changes. This paper focuses on assessing differences among pilots using the same display format. A flight simulator experiment was conducted with four pilots who had different levels of flight expertise. The analysis revealed variations in the HMM structures during the final descent segment: a 2-state HMM for the patterns of the least experienced pilot, a 3-state HMM for the two intermediate-level pilots, and a 4-state HMM for the most experienced pilot. The added "attitude-monitoring" state in the 4-state HMM reflected a flight technique well-known among experienced instrument pilots. HMM analysis methodology and interpretation issues are also discussed.