I have several datasets where motion tracking data representing a single gesture is split accross buffers with the name label just like with HHMM.
However, mubu.kNN outputs buffer index rather than the label index, so that when I load several datasets with more than one buffer for each gesture (but the same label), they get classified as different signs. Of course, it is not so important with k = 1, but since I am using a larger k, the boundaries can be somewhat modified.
Therefore, it would be nice to have an attribute that allow the kNN model to build classes from buffers with the same labels instead of different buffers.
Otherwise, I would like to know if there is a quick way to merge all buffers that share the same label into a single one; I did not find how to do it yet.