Methods and apparatuses are presented for determining whether a gesture is being performed in a sequence of source images. In some embodiments, a method includes detecting a gesture in each of one or more reference images using one or more gesture models of a plurality of gesture models. The method may also include selecting a first gesture model from the one or more gesture models that most closely matches the detected gesture, prioritizing the first gesture model over other gesture models in the plurality of gesture models for searching for the gesture in the sequence of source images, and scanning the sequence of source images to determine whether the gesture is being performed, using the prioritized first gesture model. If the gesture is being performed, the method may end scanning prior to using another gesture model of the plurality of gesture models to determine whether the gesture is being performed.