HandTracker logic and timespan between two detection

Nuitrack SDK
1

Hi everyone
I check HandTracker and I have a few questions.

  1. What is the max frequency to occur the event OnUpdate? I see that within 10 ms no more than 2 events occur.
  2. If no hand movement is detected within 5000 ms, the event subscription is reset. The event OnUpdate will not fire any more. Right? Until the next confident detection… look at the next question.
  3. The method GetProcessingTime returns the time in seconds (you need to correct the documentation, not ms) between two consistent hand movement detections. I can see that this time is never more than 5. Obviously, in the first detection event, this time should be 0 or close to it (but I couldn’t check it in debug mode). Is everything right?
2

Hi @Sergiy

  1. Frequency of the HandTracker is bound to the SkeletonTracker if SkeletonSupport option is true inside nuitrack.config (by default it’s true), otherwise it’s bound to the UserTracker.
    Since these modules work in real time, expected update frequency is about 33ms (assuming that sensor works at 30fps). Actual update rate may differ from the calculated one due to the asynchronous nature of the Nuitrack
  2. No, the triggering of the event should not depend on the movement of the hands. The event is triggered during stationary positioning of hands or even if there was not hands detected at all.
  3. GetProcessingTime returns time which was spent by HandTracker on hand recognition from the moment it received the input data to the moment it issued the result. It does not take into account the time that was spent getting its input data. Therefore, it works correctly, but I understand your confusion here.
3

Hi @a.bragin
Thank you for your reply. Time intervals are important. I plan to delve into this topic. Because I found a problem: In my test project (C#), the recognition of hand movement runs smoothly, however, if you increase the load on the system (in my case, this is the Game Maker to which the nuitrack library is included), I observe catastrophic losses of hand movement data. I’m not talking about it being a nuitrack library problem. But I need to find workarounds on how to approximate data and eliminate failing results. I’ll talk about it :-).

4

Great! Let me know if you have any other questions.

5

Closing this topic. If you have follow up notes/questions, please create a new one.

6