Research On Two-hand Gesture Recognition Based On MEMS Sensors

With the development of science and technology, the topics and application of intelligent man-machine interaction is continuously increasing. In the meanwhile, since the application occasions getting more and more complicated, developing and upgrading of man-machine interaction is expected. Interaction technology based on gesture recognition has become a research hot spot for a long time. Gesture recognition based on machine vision is sensitive to ambient light and the devices are difficult to disposal, while gesture recognition based on one-hand movements shows the shortcoming in gesture flexibility and user friendliness.Considering these problems, designing smart interaction framework and devices with features of good mobility, robustness and controllability is a very meaningful subject.After studying the actual demands and current research achievements of gesture recognition, MEMS sensing is chosen as the supporting technology; a model of two-hands gesture is built, then detailed research is carryout,which is mainly on gesture information acquisition, attitude angles calculation and two-hands gesture recognition. The gesture sensing module and recognition algorithm are designed as a result of this research.The main works are listed as following:1) Based on studying the common dynamic gesture model and performance of several kinds of MEMS sensors, a model of two-hands gesture recognition is built, and the gesture characteristic quantity is determined.2) A micro embedded hardware platform basing on MEMS sensors is designed; algorithms of sensor error calibration and multi-sensor data fusion is proposed.3) A dynamic gesture template library is built; dynamic gesture recognition algorithm based on DTW is designed; to raise the efficiency and hit rate, algorithms of sample preprocessing and acceleration are devised.4) An auxiliary gesture template library is built and a recognition algorithm based on static gesture distinguishing and template matching is designed; to associate the left hand gesture and right hand gesture, a work scheduling model based on FSM is designed, which enables the collaboration of the gesture of both hands.Additionally, a testing and analysis environment is built, in which the validity, efficiency and hit rate are verified and revealed.Experiments and analysis indicate that the achievements of this research can commendably solve technical matters in two-hand gesture recognition, achieving a mean recognition rate of 87.3% and commendable operating efficiency. The research and design has good practical significance.