Physiological Sensors
The Brain FM system ingests signals from multiple physiological monitoring modalities, collectively referred to as PPMs (Psychophysiological Monitors). Each modality captures a different window into the operator's cognitive and physiological state.
Modality Summary
| Modality | Signal Type | Sensor Form Factor | Sampling Rate | Cognitive Relevance |
|---|---|---|---|---|
| EEG | Electrical cortical activity | Head-worn electrode cap/headset | 250–500 Hz | Direct neural correlate of workload, attention, fatigue |
| ECG | Cardiac electrical activity | Chest electrode patch | 250 Hz | HRV → autonomic stress and arousal |
| PPG / BVP | Peripheral blood volume pulse | Wrist wristband | 64 Hz | Heart rate, HRV → stress and load |
| Eye Gaze | Gaze position on display | Camera-based tracker | 60–120 Hz | Attentional allocation, situational awareness |
| Pupillometry | Pupil diameter | Same camera as gaze | 60–120 Hz | LC-NE arousal, cognitive effort |
| Speech | Acoustic vocal features | Microphone | 16–44.1 kHz | Para-linguistic stress and load markers |
Invasiveness and Practicality
The modalities vary significantly in how intrusive they are to wear and operate. EEG provides the richest cognitive information - millisecond-resolution neural dynamics - but requires a head-worn device that is uncomfortable for long shifts and conspicuous in operational settings.
Speech is the most practical modality for voice-heavy environments: it requires only a microphone already present in the environment, produces no body-worn sensor burden, and is continuously available whenever the operator is communicating.
PPG offers a middle ground: a wristband is comfortable enough for all-day wear, and while the cognitive information content is lower than EEG, it provides reliable heart rate and HRV features that correlate with stress and workload.
Multimodal Combination
Combining multiple modalities improves prediction robustness:
- EEG + PPG: Neural and autonomic signals are complementary; high workload increases both frontal theta power and sympathetic HRV markers.
- EEG + Eye gaze: Neural and attentional signals together characterise situational awareness more completely than either alone.
- Speech + EEG: Speech provides near-zero-overhead monitoring; EEG can be activated as a higher-fidelity backup during critical events.
See Multimodal Learning for the technical approach to combining modalities.
Data Availability by Modality
The Brain FM's cross-modal alignment strategy directly addresses the availability imbalance:
| Modality | Public Pre-training Data | Status |
|---|---|---|
| EEG | > 30,000 hours (TUH EEG Corpus) | Large-scale SSL possible |
| ECG | Clinical archives; moderate | SSL feasible |
| PPG | Sparse cognitive datasets | Needs cross-modal alignment |
| Eye gaze | Very sparse | Needs cross-modal alignment |
| Speech | Large general ASR corpora (not physiological) | Domain gap issue |