How to add a none option in scikit learn predict

Question

I have the next problem, I'm trying to classify one of four hand position, I'm using SVM and this positions will be used to make commands in my program, the predict function works fine, but for example if I made some other gestures (none of the 4 that I use for commands) the predict function try to classify this gesture in one of the original four, I want to know if is possible to say "this gesture is none of the four that I know".

The final function of this is similar to the behaivor of some commands of the kinect in Xbox for example, you can move your hands but the Xbox will only react to a especifict gesture.

I'm using Python, pandas, scikit-learn and leap motion, I training my network with data collected from four gestures.

Thanks in advance.

Answer 1

Instead of treating this as a classification problem with 4 levels, treat it as a classification problem with 5 levels. 4 levels would correspond to one of the original four and the 5th level can be used for all others.