AELINanalyzes a gentle morning walk with the real 33-point body pose and risk framework.

Care vision

Care vision / spatial safety

AELIN

An explainable proof that translates posture, space and time into a care signal that is always verified by a human.

Working pose proof of concept

The case in 30 seconds

AELIN

Problem: An explainable proof that translates posture, space and time into a care signal that is always verified by a human.
Solution: From fixed space camera to explainable pose, risk zone and a quiet human verification route.
Role and focus: Pose, context, care flow
Status: Working pose proof of concept
Result: Brand, product site, poseproof

Live lab / try it yourself

See when posture becomes a signal

Follow the same space throughout the day and inspect how posture, risk zone and time context together form a humanly verifiable care signal.

Open the pose proof View the story

01 / InputAttitude, space and timeYou can change or choose this yourself

02 / SystemPose estimation and context rulesPose estimation · Spatial context · Human verification

03 / EvidenceAn explainable concern signalThis makes the operation visible

InputFixed space camera

Analysis33 body landmarks

ResultGreen, orange, red

DirectionHuman verifies

Brand vision / perceiving with restraint

Care that sees sooner

The protective arc embraces a human presence signal. Pine green stands for peace and professional direction; coral only appears when a healthcare team really needs to look.

The design question

When does standing still become a signal?

Not every lying or still body is an emergency. The product value lies in the combination of attitude, location, change and duration, and in a quiet route to human verification.

Starting point

Can one fixed camera make a possible fall visible earlier without turning every moment of rest into an alarm?

Attitude

Upright, sitting or horizontal

Pose keypoints make body posture measurable without facial recognition.

Context

Sofa is not a floor

The same posture takes on a different meaning in a rest zone or free floor zone.

Border

A signal is not a diagnosis

AELINcan mark a possible trap. It cannot prove from one image that someone is in cardiac arrest, asphyxiation or other medical incident.

One home / four moments

Context about the day

The same camera, occupant and room deliberately keep the test small. Pick a moment and see the real locally calculated pose overlay.

GREEN · pattern matching

Upright and on the move

33 visible landmarks, torso 2.6° off vertical and a high body center. No floor position.

AttitudeVertical

ZoneWalking route

Pose visibility95,7%

ContinuedObserve

Working first test

Four fixtures classified correctly

The first run failed meaningfully: in the wide view, an armchair could resemble a human. Overlapping camera tiles made the resident larger and then allowed the model to choose the anatomically strongest pose in all four scenes.

MediaPipe pose overlay highlights the supine resident in red

4/4expected fixture labels in this little test

Which makes image 4 red

All 33 landmarks are visible, the torso is 59.1° from vertical and the body center is low in a calibrated floor zone.

Which this doesn't prove

No sensitivity in the real world, no passage of time and no distinction between all possible causes of lying on the floor.

pose + zone + time → signal → human verification

Open source route

From frame to action

The proof uses MediaPipe. For real video, the same product layer can later be powered by Ultralytics, MMPose, and a temporal action model such as PoseC3D or ST-GCN.

Find person

Overlap tiles prevent the resident from remaining too small in a wide room camera.

Measure pose

Keypoints translate body position into explainable geometry.

Track time

Center drop, change in orientation and duration distinguish falling from bending.

Understanding space

Bed, sofa, chair and floor each have their own context.

Enable people

Call, speak, watch or send care according to the location protocol.

Safety boundary

Not everything is visible

Choking, shortness of breath, or loss of consciousness on the couch can look like normal sitting. This requires time, micro-movement, audio, an active speech check or additional sensors.

Clearly visible