I have an ANBC classifier working for gesture detection, in a 4 class problem (4 different gestures).
The classifier is working well except for one problem: When no gesture is performed, it outputs a result (two out of the four gestures are "close" to the default position), despite of high distances to the class given as a result.
I have enabled the NullRejection feature (anbc.enableNullRejection(true)) and also tested different coefficients, even 0, with no success.
Does anyone have a clue about this? I have tried to filter the results manually attending to the distances to the class given as a result, but it is not a good solution...
Hi all;
I have an ANBC classifier working for gesture detection, in a 4 class problem (4 different gestures). The classifier is working well except for one problem: When no gesture is performed, it outputs a result (two out of the four gestures are "close" to the default position), despite of high distances to the class given as a result. I have enabled the NullRejection feature (anbc.enableNullRejection(true)) and also tested different coefficients, even 0, with no success.
Does anyone have a clue about this? I have tried to filter the results manually attending to the distances to the class given as a result, but it is not a good solution...
Best, Alberto