so, through k-means we have to pass all the points at once, but do we need to be able to distinguish between what are the gestures (ergo, feed them one by one) to HMM learning process?