Segmentation and recognition of continuous gesture based on chaotic theory.

The performance of conventional continuous gesture recognition algorithms is mainly affected by factors such as incomplete keyframe detection, input of unconscious gestures, or variations in duration and movement range. The hypothesis is made that the irregularly sampled data of continuous gestures can be approximated with a particular type of dynamic system, and the characterization of these nonlinear dynamics will help with the trajectory partition and establishment of feature vectors. Finally, the proposed algorithm is evaluated with a database of alphabetic gestures, and the experiment results indicate that our framework has a high recognition rate of around 93.6% while maintaining its performance in the segmentation of continuous gestures. [ABSTRACT FROM AUTHOR]