The Research Of The Hmm Dynamic Gesture Recognition Based On Trajectory Feature

With the development of human-computer interaction, people’s vivid and intuitive hand gestures are treated as human-computer means of communication, which can get rid of the bondage of traditional input methods such as keyboard and mouse of the users. The development has immeasurable significance in the human-computer interaction. Form the gesture input device classification, there are two types of the main research directions of hand gestures, based on the data glove hand gesture study and the visual hand gesture study. Form the hand gesture recognition object classification, the hand gesture recognition research is divided into static hand gesture recognition and dynamic hand gesture recognition.The work of this paper is mainly expounded from two aspects:based on the HU invariant moment feature static hand gesture recognition research and the dynamic hand gesture recognition based on the hand gesture trajectory.In the project of the real-time gesture modeling and recognition technology research in the medical virtual-reality interaction, the hand gesture recognition is the key technologies to realize the virtual-reality interaction. The accuracy and stability of the current hand gesture methods are still unsatisfactory due to various reasons. The paper proposes a hand gesture recognition method based on HU moments. The HU moments have invariance property of translation, rotation and size. By extracting HU moments of the hand contour, this paper applies support vector machine (SVM) method to achieve the real-time hand gesture recognition. Experiments prove that the proposed method has very good precision and stability in virtual reality interaction.Compared to static hand gesture recognition, the dynamic hand gesture recognition is not only the hand shape change but also including the changes in the space and time. The dynamic hand gesture recognition research relates to the spatio-temporal trajectory. The paper proposes dynamic hand gesture tracking method based on the improved particle filter. The paper also proposes that the direction angle of the dynamic hand gesture’s trajectory characteristic as the feature. This paper applies Hidden Markov Model (HMM) to achieve the dynamic hand gesture recognition. The experiments prove that the method of the dynamic hand gesture recognition based on the spatio-temporal trajectory characteristic of the hand action having a good experimental effect.