Today's driver monitoring systems are almost all cameras. Eye gaze, head pose, blink rate, facial orientation — vision-based sensing dominates because it's non-contact, relatively cheap, and doesn't require the driver to wear or touch anything special. But camera-based monitoring has a well-known weakness: it infers driver state indirectly, from external behavioral cues, and those cues degrade exactly when conditions get difficult — poor lighting, glare, obstructions, or a driver simply not looking at the camera.
A growing body of automotive research is exploring a complementary approach: sensing the driver's actual physiological state — heart activity, brain activity, muscle activity — directly, using sensors built into the steering wheel, seat, and safety belt rather than a wearable device. This is squarely the kind of problem our driver monitoringwork touches, so here's what physiological sensing in the cabin actually involves.
Why physiological signals, specifically
Camera-based systems observe behavior — how a driver looks, moves, and orients their head — and infer state from that. Physiological sensing measures the underlying state more directly. Behavioral signals alone lack the sensitivity to catch changes in a driver's underlying health status, while physiological monitoring can objectively assess condition in real time and potentially catch early signs of a medical emergency — not just drowsiness. That distinction matters for edge cases cameras handle poorly: a driver who looks alert but is in the early stages of a cardiac event, or whose eyes are open but whose heart rate variability signals genuine physiological fatigue rather than momentary distraction.
Where the sensors actually go
Since drivers generally don't wear dedicated medical sensors behind the wheel, researchers have focused on embedding sensing into surfaces the driver is already touching or resting against.
- The steering wheel.One of the most studied approaches embeds ECG sensing into the wheel rim using electrically conductive fabric electrodes that read signal from the driver's palms. In one real-world test, this sustained ECG acquisition across a two-hour drive, classifying driver state as normal, fatigued, or drowsy; a related study found roughly 90% of measured heartbeat intervals had error below 5% against a reference ECG. The core engineering challenge is maintaining constant skin contact — acquisition generally needs both hands on separate conductive zones simultaneously, which real driving behavior doesn't always guarantee.
- The seat and safety belt.Capacitive electrodes embedded in the seat offer a heart-rate path that doesn't depend on hand position at all. Ballistocardiography uses strain-gauge sensors in the seat or belt to detect the mechanical signature of cardiac and respiratory activity — though vehicle vibration during actual driving makes this a genuinely difficult signal to isolate cleanly.
- Non-contact camera-based sensing.Cameras themselves are being extended beyond behavioral observation into physiological sensing, using camera-based PPG to extract heart rate from subtle color changes in the driver's face — blurring the line between "camera-based" and "physiological" monitoring rather than treating them as separate categories.
- Multimodal steering-wheel platforms. More comprehensive research prototypes combine ECG, EMG (muscle activity), EDA (skin conductivity), PPG, and hand-pressure sensing into a single retrofittable add-on — explicitly designed to install on existing vehicles rather than requiring a new wheel or seat.
Fusing that many simultaneous signal chains cleanly is exactly the kind of problem we approach through electrical engineering and AI integration together — the sensing modality choices mirror the same framework we use for medical biosensor selection, just applied to a noisier, in-cabin mounting surface.
The trade-offs that keep this from being a simple swap
Physiological sensing in the cabin isn't a strictly better replacement for cameras — it comes with its own real constraints. Signal quality is inherently harder to guarantee than with a wearable, since a steering wheel or seat sensor can't be positioned or tensioned the way a wrist strap can. Motion and vibration are a bigger artifact source than in wearables — the same motion-artifact challenges we cover in our wearable cardiovascular sensing work are amplified in a vehicle, where the whole cabin vibrates and hands move on and off the wheel constantly. And contact isn't guaranteed — a steering-wheel or seat sensor only works when the driver happens to be touching the sensing surface in the right way, a limitation cameras don't share. The realistic framing is complementary, not competitive: cameras for continuous behavioral observation, physiological sensors for the moments and metrics cameras genuinely can't capture well.
What this means for suppliers and cabin designers
- Design physiological sensing as an addition to camera-based DMS, not a replacement — fusing vision, physiology, and vehicle behavior is the direction with the most real-world robustness.
- Treat contact reliability as a first-class design constraint, the way sensor placement is treated in wearable design — a steering-wheel ECG system needs a fallback for the significant fraction of driving time when both hands aren't positioned correctly.
- Borrow directly from wearable signal-processing experience — the artifact-rejection and motion-compensation techniques developed for wrist- and ear-worn PPG/ECG sensing translate directly, even though the mounting surface is completely different. This is where our firmware engineering work on real-time sensor fusion applies just as much in-cabin as on the wrist.
- Consider retrofittable, add-on architectures for near-term deployment rather than requiring a full steering wheel or seat redesign that only new vehicle platforms could adopt.
Bringing it together
Physiological sensing is entering the vehicle cabin because behavioral observation alone — however sophisticated the computer vision gets — has a ceiling on what it can infer about a driver's actual internal state. Steering wheels, seats, and belts are emerging as genuinely viable physiological sensing surfaces, carrying over much of the same PPG and ECG expertise developed for wearables into a noisier, more constrained mounting environment. The likely future isn't physiological sensors replacing cameras — it's the two working together.
Exploring physiological sensing for a cabin or wearable platform? Talk to CoBuild Labs — or see related sensing builds in our Work portfolio.