PlatformsFaststream SiliconFaststream RadioFaststream VisionConnected EdgeFaststream SecureMobility & Rail
ProductsSemiconductor IPWireless & RANEdge & GatewaysTracking & IdentificationSoftware & FrameworksConnected Systems
Technology5G protocol stackWireless and RF architectureBaseband and low PHYForward error correctionControl and data planeHigh-speed interfacesFirmware and bootSilicon root of trustSoftware-defined vehicleAutomotive OTAFunctional safety
AIAI Engineering ServicesEdge AI & Embedded MLComputer Vision EngineeringSensor Fusion & PerceptionAI Silicon & AccelerationMLOps for DevicesAI Visual InspectionPredictive MaintenanceDriver MonitoringVideo Analytics & Safety
SolutionsSemiconductorIndustrial AIConnected ProductsAsset TrackingBluetooth AoA RTLSWearable TrackingAutomotive & MobilitySmart InfrastructureSecure IdentityWireless & SatellitePrivate 5GSmart WashroomsFuel ManagementSmart BuildingsWorker SafetyEnergy MonitoringSmart AgricultureSmart CityAutonomous PlatformsAssembly AutomationLiDAR Rail SafetyHardware Wallet
IndustriesSemiconductorTelecommunicationsIndustrial & ManufacturingAutomotive & MobilityTransportation & RailAerospace & DefenceHealthcare & MedicalEnergy & UtilitiesOil & GasRetailConsumer ElectronicsMedia & EntertainmentSmart Infrastructure & IoT
ServicesSystem Integration overviewASIC & SoC DesignRTL to GDSIIVerification methodologyDFT and silicon testLow-power designMixed-signal integrationDesign enablementFPGA DesignFPGA-to-ASIC ConversionAnalog, Mixed-Signal & RFHardware & High-Speed PCBEmbedded SoftwareCloud, OTA & Device ManagementManufacturing TransitionHow we engage
CompanyAbout FaststreamEngineering ExcellenceLeadership & OrganisationHow We EngageQuality & ComplianceStandards & EcosystemPartners & EcosystemTrust CentreLocations & DeliveryNewsroom & MediaCareersCase StudiesKnowledge CenterWhite PapersGlossaryNewsletterResources & Support
ContactStart a projectHow we engage
Talk to us
CONNECTED PRODUCT

Warning a tired driver is easy. Being right often enough to be trusted is not.

A fatigue monitor earns its place by being right — a system that cries wolf is taped over by lunchtime. Detecting eye closure and distraction from an in-cab camera is only useful if it holds across every face, sunglasses, night driving and vibration, and warns early enough to matter without nagging.

DomainConnected product, driver monitoring
PlatformsConnected Edge
ScopeIn-cab imaging to graded alert
Binding constraintRight often enough to be trusted, not just sensitive
DisclosureRepresentative programme; customer not named
CONTEXT

Where this started.

A driver-monitoring system watches the cab with a camera and infers drowsiness and distraction from eye closure, gaze, head pose and blink dynamics. The value is a timely warning; the risk is a system so prone to false alarms that drivers disable or ignore it.

The difficulty is variation. Faces differ, sunglasses and clear glasses reflect, night driving means infrared imaging, and a cab vibrates and swings between glare and shadow. A model tuned in a lab drifts the moment it meets that range, and every false alarm erodes the trust the system depends on.

So it is engineered for the cab it lives in: infrared imaging that works day and night, models robust across faces and eyewear, and a graded alerting logic that distinguishes a genuine microsleep from a glance at a mirror.

CHALLENGES

4 problems, named.

Stated before any of them had an answer.

01

False alarms kill trust

A monitor that warns at every mirror-check or shadow is switched off or ignored; the whole value rests on a low false-alarm rate.

02

Faces and eyewear vary hugely

Different faces, sunglasses, clear glasses and reflections defeat a model that learned a narrow range; robustness across people is the requirement.

03

Day and night, glare and shadow

A cab swings from bright glare to darkness; the imaging has to work across all of it, which usually means infrared, not visible light alone.

04

Early enough to matter

A warning that fires as the eyes are already closed is too late; the system has to read the onset of fatigue, not just its arrival.

ARCHITECTURE

How it was built.

DRIVER MONITOR, TRUSTED ALERTSSEEIR imagingDay, night, through glassesFace and eye trackingAcross faces and poseRobust featuresBlink, gaze, headINFERFatigue modelOnset, not just closureDistraction modelGaze off-roadConfidenceSuppress the marginalALERTGraded warningEscalates with certaintyEdge inferenceIn-cab, no cloudEvent loggingFor review, not blame

The model is the easy half. Working across every face, in glare and darkness, and warning early without nagging is what turns a demo into a device a driver leaves switched on.

CONTRIBUTION

What Faststream did.

The scope of the work, rather than a capability list.

WHAT WAS HARD

The parts that consumed the schedule.

Rarely the subsystem that sounds difficult.

01

Suppressing false alarms

The central engineering is not detecting closed eyes but not warning on the thousands of harmless glances and shadows; that is where the trust is won or lost.

02

Generalising across people

A system that works for one face and not another is useless in a shared fleet; robustness across the population is hard and essential.

03

Imaging a moving cab

Glare, darkness, vibration and eyewear reflections all attack the imaging; infrared and careful optics are needed before any model helps.

04

Reading onset, not arrival

Catching fatigue early enough to matter means modelling its onset dynamics, which is subtler than detecting a fully closed eye.

OUTCOME

What resulted.

Alerts drivers trust

A low enough false-alarm rate that the system stays switched on and is acted upon.

Works across the fleet

Reliable across faces, eyewear and lighting, not just the development set.

Day and night

Consistent detection in glare and darkness through infrared imaging.

Immediate and private

Edge inference in-cab, warning without a cloud round-trip and without shipping video off the vehicle.

Confidentiality

Customer projects are presented at property, capability, outcome and integration level. Customer names, internal architecture, confidential deliverables and commercial terms are not disclosed. Where a detail would identify a customer it is omitted rather than approximated. More is available under a non-disclosure agreement, within the limits the customer has agreed.

PRODUCTS AND CAPABILITY USED

What this was built from.

Every item links to its own page.

WHERE THIS APPLIES

Industries this serves.

COMMON QUESTIONS

Questions this programme gets asked.

01

Why is a fatigue monitor hard if detecting closed eyes is easy?

Because the product is judged on trust, not on detection. Recognising a closed eye in a clear image is straightforward; doing it reliably across every driver, through sunglasses and clear glasses, in glare and darkness, in a vibrating cab, and warning early without firing on every harmless glance is the real problem. A system that raises false alarms gets disabled or ignored, so suppressing them matters as much as catching genuine fatigue.

02

Why infrared rather than a normal camera?

Because a cab swings between bright glare and total darkness, and a visible-light camera fails at night and struggles with sunglasses. Infrared imaging works across day and night and sees the eyes through many sunglasses, giving the model a consistent input regardless of ambient light. Getting that imaging right is a precondition for the detection working at all.

03

Does the video leave the vehicle?

No — inference runs on edge hardware in the cab, so the fatigue and distraction decisions are made locally and immediately, and raw video does not need to be streamed to the cloud. That keeps the warning instant, avoids depending on connectivity, and keeps the driver's image private, with only events logged for review rather than continuous footage.

FOUND THIS USEFUL?

Pass it on.

Written for engineers. Share it with one.

ShareLinkedIn
KEEP READING

Related work.

BUILD WITH FASTSTREAM

Bring us the difficult part.

Tell us the specification, the constraint and the deadline. Programmes that cross silicon, radio, embedded and AI are where Faststream is strongest.