Azure_Kinect_sensor-README

Sensor - Azure skeleton

Live 3D positions of body joints from the Azure Kinect sensors.
Choose the joints you need, each gets its own outlet.

In: none · Out: one CHOP per joint (tx ty tz)

Add joints (+/-) and pick a joint in each slot (head, hand_l, hand_r, pelvis, ...). The outlets update by themselves.
Set Person: 1 follows the first tracked person (p1), 2 the second, and so on.
Hang a null named after the joint off each outlet (null_head, null_handl, ...) and work from those.
More people? Copy the component and give each copy its own Person.
Not in the lab? Turn on Mouse and keep building.

Mouse: off = sensor data, on = all joints follow the mouse
Mouse orientation: which plane the mouse drives (XZ = across the floor)
Person: which tracked person this copy follows (1-8)
Add joints: adds or removes joint slots
Configure: restarts the sensors on the Kinect host. Only use it when tracking hangs. Greyed out in copies.
Fallback origin + Position: where joints rest when there is no data. Off = 0 0 0.

All values at 0 (or at Position): not in the lab, nobody with that Person number in the room, or no data for 0.5 s. Turn on Mouse to keep working.
Mouse orientation greyed out: Mouse is off.
Configure greyed out in a copy: only the first copy talks to the host. The other copies read its stream through a blue Select called shared_stream.
No fps in a copy's view: only the first copy receives the frame rate.
Someone gets a new number after leaving and coming back: numbers follow tracks, not individuals.
A person returning snaps to their position instead of gliding: on purpose.
Hands are a bit less steady than head and pelvis: the sensors are less sure about hands. Wrists are steadier and move almost the same.
Left and right hand swap when someone turns their back: a known sensor limitation.
Don't save a copy as the .tox. Save the original only.