Driver Monitoring System

A driver monitoring system estimates whether the driver is attentive and capable of performing the driving task. Monitoring is especially important with sustained assistance, but sensing the driver and controlling the vehicle are separate functions.

Sensors used

Driver monitoring normally uses an interior Cameras, often with near-infrared illumination described in Infrared and thermal cameras, to estimate gaze, eyelid state and head pose. Steering-wheel touch or torque and other control inputs can provide narrower supporting evidence through Positioning and Vehicle Motion.

The system must manage occlusion, eyewear, seating position, lighting and demographic performance. See Sensor Fusion for combined confidence, Calibration, Cleaning and Sensor Health for degraded sensing and How to Validate Sensor and Perception Claims for evaluating monitoring claims.

Direct and indirect monitoring

Direct monitoring normally uses an inward-facing camera, often with infrared illumination. Depending on the implementation, it can estimate gaze direction, head pose, eyelid opening, blink behavior and whether the face or eyes are visible.

Indirect monitoring infers engagement or impairment from driving behavior. Inputs can include steering corrections, lane position, trip duration and control use. Steering-wheel torque, touch or capacitance sensing indicates interaction with the wheel; it does not prove that the driver is watching the road.

Different driver states

A credible system should distinguish states that require different responses:

a single long glance away;
repeated short glances that accumulate into distraction;
phone use;
increasing drowsiness;
microsleep or sleep;
impairment not caused by fatigue; and
an unresponsive driver.

Euro NCAP’s 2026 Driver Engagement protocol evaluates transient distraction separately from non-transient conditions such as drowsiness, microsleep, sleep and unresponsiveness. It also considers lighting, eyewear, facial features and partial sensor blockage.

Warnings and intervention

The first response is normally a visual warning combined with an audible or haptic alert. A vehicle may then increase warning intensity or change the sensitivity of lane and forward-support systems.

If the driver remains unresponsive, a separate emergency function may maintain lane position, increase following distance, reduce speed or stop the vehicle. More advanced systems may move toward a slower lane or emergency area where regulations, road layout and sensor capability allow. These actions depend on steering, braking, environment perception and hazard signaling; the DMS camera does not perform them by itself.

Monitoring during Level 2 assistance

Hands-free Level 2 still requires eyes-on supervision. A direct monitoring camera can check visual attention more meaningfully than steering-wheel torque alone. The assisted-driving system should communicate its active state clearly and respond when monitoring becomes unavailable.

Level 3 creates a different role because the automated driving system performs the driving task while active. The distinction is covered in Level 2 vs Level 3, while fallback and safety architecture are covered in Safety, Redundancy and Regulation.

What buyers should check

whether monitoring is camera-based, indirect or both;
whether gaze and phone use are detected;
performance with sunglasses, darkness and partial occlusion;
when warnings begin and how they escalate;
whether assistance is disabled when monitoring is unavailable;
the response to microsleep, sleep and unresponsiveness; and
whether emergency-stop claims apply to the exact market and software.