Fatigue
Fatigue in cognitive monitoring refers to the progressive reduction in alertness, processing speed, and cognitive efficiency caused by sustained mental effort or sleep deprivation. It is distinct from workload (which can be high without fatigue) and from stress (which involves arousal rather than depletion).
Why Fatigue Matters
Many operational roles require sustained cognitive performance across rotating shift patterns, including night shifts that cut across the natural circadian sleep phase. Fatigue accumulates:
- Over the course of a shift (time-on-task fatigue).
- Over successive days of early or night shifts (sleep debt accumulation).
- Over the length of a career for individuals with chronically disrupted sleep patterns.
Microsleeps - brief, involuntary lapses into a sleep-like state lasting 1–15 seconds - can occur in severely fatigued individuals without subjective awareness. A 10-second microsleep during a critical operational situation can be catastrophic.
Even sub-clinical fatigue levels (before microsleeps occur) produce measurable reductions in sustained attention, slower response times, reduced vigilance for rare events, and impaired error detection.
Physiological Correlates
EEG
Fatigue produces characteristic changes across multiple frequency bands:
| EEG Feature | Change with Fatigue | Neural Basis |
|---|---|---|
| Alpha power (8–13 Hz) | Increases (opposite to workload!) | Reduced arousal, cortical deactivation |
| Theta power (4–8 Hz) | Increases | Drowsiness, reduced executive engagement |
| Beta power (13–30 Hz) | Decreases | Reduced active processing |
| Alpha/beta ratio | Increases | Strongly correlated with subjective sleepiness |
| Theta/alpha ratio | Changes | Reflects transition toward sleep |
| Frontal alpha asymmetry | Shifts toward reduced left frontal activity | Withdrawal, reduced motivation |
Alpha behaves differently in fatigue vs. workload
Alpha power increases with fatigue but decreases with workload (alpha suppression). A model trained primarily on workload datasets may confuse the two states if alpha features are used naively. Careful feature design or joint multi-state training is required to disambiguate.
Ocular Measures
Eye-based fatigue signatures are among the most reliable:
| Measure | Change with Fatigue |
|---|---|
| Blink rate | Initially decreases, then increases as fatigue becomes severe |
| Blink duration | Increases; slow, heavy blinks |
| PERCLOS (% eye closure) | Increases; proportion of time eyelids are 80%+ closed |
| Saccade velocity | Decreases; slower eye movements |
| Fixation stability | Decreases; microsaccades increase |
PERCLOS is the primary operationally validated fatigue indicator used in driver monitoring systems.
Heart Rate and HRV
- Heart rate decreases slightly with fatigue (reduced sympathetic tone).
- HRV in the LF band may increase as parasympathetic recovery dominates.
- HRV patterns shift toward lower frequency as sleep deprivation accumulates.
Speech
- Speech rate slows and becomes more monotone.
- F0 variability decreases (flatter, less expressive prosody).
- More hesitations and disfluencies.
- Voice tremor may increase with severe fatigue.
Fatigue Assessment Instruments
| Instrument | Format | Application |
|---|---|---|
| Karolinska Sleepiness Scale (KSS) | 9-point scale (1 = extremely alert, 9 = extremely sleepy) | Repeated during shift to track accumulation |
| Borg Fatigue Scale | 6–20 point scale | Physical fatigue; used in WAUC |
| PANAS | Positive/negative affect | Captures low-activation negative affect correlated with fatigue |
Time-on-Task Fatigue vs. Sleep Deprivation Fatigue
These two fatigue types have different physiological signatures:
| Type | Cause | Timescale | Primary Signature |
|---|---|---|---|
| Time-on-task | Sustained mental effort without break | 30–90 minutes | Theta ↑, sustained attention ↓ |
| Sleep deprivation | Insufficient sleep | Hours to days | Alpha ↑, PERCLOS ↑, circadian misalignment |
Operators are vulnerable to both: time-on-task fatigue within a single session, and sleep deprivation fatigue across shift rotations. The Brain FM must distinguish both from baseline and from each other.
Fatigue in the Monitoring Pipeline
For operational fatigue monitoring, the system tracks:
- Short-term fatigue indices from the current sector session (30–60 minute rolling window of alpha/theta features).
- Cumulative shift indicators that track fatigue accumulation across the full shift duration.
- Trend prediction - projecting current fatigue trajectory to estimate when degradation will reach a critical threshold, enabling proactive rest break scheduling.
Unlike workload (which can vary rapidly), fatigue accumulates gradually and is best monitored over longer temporal windows. The BFM's sequence length agnosticism enables long-window analysis alongside short-window predictions.