关键词: 手势识别, 虚拟现实交互, 混沌动力学, 特征工程

Abstract: For the problem that the performance of conventional continuous gestures recognition was mainly affected by factors such as variations in the movement duration and inaccurate feature extraction,the nonlinear dynamic model of chaos theory was used to explore the inherent chaotic motion characteristics of volley dynamic gestures,and a fuzzy interactive framework based on chaotic dynamic characteristics was established.Leapmotion head-mounted display was embedded with discrete observation to obtain finger tracks.Assuming that a particular type of chaotic dynamical system might simulate the gesture motion model,the feature matrix composed with chaotic feature factors was established based on Phase Space Reconstruction (PSR) for identifying classification.The hypothesis was evaluated with a database of alphabetic gestures,and an accuracy of 96.6% in the experiment was achieved.

Key words: gesture recognition, virtual reality interaction, chaotic dynamics, feature engineering