Extending system awareness to the operator
Intelligent systems continuously read speed, position, attitude, temperature, energy level and communication quality. The operator's condition is rarely treated as part of the same information chain. Driving, flight operations, industrial work and long-duration monitoring share a common risk: the equipment may remain healthy while the person is no longer in an appropriate state to continue the task. Fatigue, reduced attention, gaze deviation and slower reactions accumulate over time and change the quality of human–machine coordination.
L-IN EYE explores how eye, facial and operational cues can become continuous system inputs. The work considers eyelid closure, blink rhythm, gaze direction, head pose and their development over time. A single action has limited diagnostic value; a temporal sequence provides a more stable indication of state.
Perception pipeline
Human-state monitoring must remain usable under changing illumination, head movement, eyewear and rapid gaze shifts. The system therefore begins with imaging conditions, combining near-infrared illumination, narrow-band filtering and camera placement to improve visibility around the eyes. The processing chain then locates facial and eye regions, extracts fatigue- and attention-related features, and forms a state estimate across a time window.
The pipeline can be described as imaging, localisation, features, temporal analysis, state and feedback. Each layer should preserve confidence information so that higher-level systems can understand the basis of a decision. The resulting state can support driver assistance, flight-task management or industrial safety interfaces.
Designing for multiple contexts
Driving emphasises road attention and fatigue trends. Aviation places greater weight on attention allocation during long missions. Industrial operation also depends on the workstation, equipment state and procedure. These contexts can share a sensing foundation, while thresholds, time windows and feedback policies remain application-specific.
L-IN EYE therefore follows a platform approach: the lower layer stabilises acquisition, the middle layer analyses features and time series, and the application layer organises feedback and records. Future versions can combine workload, environmental information and equipment state to show how operator condition relates to operational risk.
Next steps for human–machine coordination
When human state enters the system model, automation can detect changes in coordination quality earlier. It can provide a gentle prompt during the initial decline in attention, adapt interaction as risk rises, and hand state information to a higher-level safety function when required. The design priority is clear interpretation, controlled false alarms and transparent communication with the operator.
Further validation is required across larger datasets, lighting conditions and user groups. The long-term direction remains consistent: future intelligent machines need to understand environment, task and person together. That shared loop is the basis for more dependable automation.