The "end" time is hard to interpret (one has to subtract); it would be better to print duration (= end - start) instead.
Also, AdVent engine assumes that recognition runs for no longer than i + REC_DEADBAND time, where i is recognition interval parameter and REC_DEADBAND is a hardcoded constant (before issue #24 is resolved). So it would be good to report if de-facto recognition duration exceeds forecast by a good margin (say, by more than 0.5 s), as it could affect AdVent efficiency.
Currently we have
-l debug
producing something like this:The "end" time is hard to interpret (one has to subtract); it would be better to print duration (= end - start) instead.
Also, AdVent engine assumes that recognition runs for no longer than
i + REC_DEADBAND
time, wherei
is recognition interval parameter andREC_DEADBAND
is a hardcoded constant (before issue #24 is resolved). So it would be good to report if de-facto recognition duration exceeds forecast by a good margin (say, by more than 0.5 s), as it could affect AdVent efficiency.