Our system works with an infinite input space, namely images of human motions. Between the system not working at all and working optimally, there is a big space where it works sometimes. This space is difficult to observe and control. In fact, the best observers are the user themselves. Only they know if the result fulfills the expectation.
With the session quality, we want to provide a hint to end user. They are able
to observe if the system works. If they decide it does not work good enough, the
session quality provides a guideline what can be improved. The API provides
qualities for the two subjects
latency: Describes the round trip time for an image. If the latency is high, the system has less frames per second to analyze a motion, less time resolution. This means, important states could be missed.
environment: Summarizes the lighting, contrast and clarity of an image. The lower the quality, the less possible it is to recognize the human in the image.
The vague wording should show you, that we cannot give an absolute statement. The system can also work in a bad environment and with high latency, but the chances are lower. The following four qualities are defined:
none: No quality information available yet.
bad: The quality may impair the movement analysis. We cannot guarantee a good performance.
ok: The quality is good enough to perform the movement analysis, but the accuracy can be affected.
good: The quality is optimal and enables the movement analysis.
Important Note: While we managed to determine an adequate estimate how the latency affects our motion analysis, the environment quality must be viewed with caution. We didn't yet find the exact relation between the environment and our systems performance, which is why the environment quality should be considered as auxiliary information.