The MGestr system reads an multitouch input stream to identify gesture segments and subsequently classify them using the currently active gesture set. A gesture segment is a sequence of frames of a fixed number of fingers moving accross the multitouch surface. The start of a gesture is made by a finger_down event, by placing a finger on screen. If more fingers are placed on screen, the previous frames are discarded. Once the finger(s) begin moving, the frames are stored as a gesture segment. To identify the end of gestures, two delimiters are used.

The first delimiter used is a when a finger is lifted off the surface. The gesture segment is marked as ended.

The second delimiter checks whether all the fingers have stopped moving (within a small threshold). If all fingers have stopped moving, the gesture segment is ended. Any subsequent movement is considered a new gesture segment.

To filter noisy input, a minimum number of frames are required to be collected before a gesture segment is classified.

The gesture segment is formed, it is tested against the active gesture set. The result of the classification is sent via OSC messages on the output port.