Research On Gesture Recognition Technology Of Quadruped Robot In Moving State

The movement flexibility of the quadruped robot has demonstrated absolute advantages in the field and other complex environment operations.Good interaction methods will greatly improve the maneuverability of quadruped robots.Compared with interaction methods such as voice recognition,face recognition,and brain-computer interaction,gesture interaction has intuitive actions and strong anti-interference capabilities.According to the functional requirements of the quadruped robot,the interactive gesture set is usually designed to have both static and dynamic mixed gesture sets.Therefore,the design of mixed gesture sets and the recognition method for mixed gesture sets are the technical difficulties faced by the realization of the gesture interaction function of the quadruped robot.At present,the research on gesture recognition focuses on the relatively still state of the human body,and there is little research on non-stationary conditions such as the operator's travel.In response to the above problems,around the research goal of realizing the interactive control of gestures in the quadruped robot,this paper completes the research work based on the gesture collection system built by inertial sensors as follows:(1)Model construction and motion elimination in the process of human movement.With reference to the common hand models in the past,based on the analysis of hand structural characteristics and motion constraints,a simplified hand model is proposed.Analyze human movements and establish human movement models.Human body motion state is measured by an inertial sensor,and the human body motion is eliminated by coordinate conversion based on relative kinematics.(2)Construction of static and dynamic interactive gesture set for quadruped robot.Aiming at the interactive control requirements of the quadruped robot during the process of travel,a gesture interaction instruction set is designed.According to the actual application needs of the quadruped robot,the gesture set is further optimized,and finally the gesture interaction gesture set during the traveling process of the quadruped robot is designed.(3)Research on Gesture Recognition Algorithm Based on Mixed Gesture Set.Propose a two-layer gesture recognition classifier based on extreme learning machine and dynamic time warping theory.Inertial sensor data after data preprocessing,extract corresponding feature vectors and use corresponding classifier to realize gesture classification.For static gestures,a mixed kernel function extreme learning machine based on extreme learning machine theory is used for gesture recognition.For dynamic gestures,template matching is achieved through multi-template template library combined with KNN algorithm.In addition,the template update algorithm is proposed to improve the adaptability of the algorithm.(4)Gesture recognition experiment and analysis.The experimental platform used in this paper is introduced,and gesture recognition experiments and analysis are conducted to verify the effectiveness of the motion elimination scheme and gesture recognition algorithm in this paper.