As Human-Machine Interaction evolves with new technologies, more and more sensor data needs to be processed in a short period of time. To look at the right spots in data for dynamic mid-air gesture classification becomes a key factor.
Therefore, we present a novel method for frame by frame gesture segmentation with the goal of giving immediate operator feedback during the execution of gestures and assist the operator in a natural way, also the adjustment to individual operators or scenarios are possible due to the direct approach. That is realized by using a support vector machine combined with a moving average filter to generate complete segments and conduct pre-classifications as a by-product. This setup has no limitations in flexibility concerning gesture execution duration because of its direct approach.
For the full paper, please contact me