Research On Gesture Recognition Control System Based On Wheeled Inverted Pendulum Robot

With the progress of the computer era,intelligent robots appear in people's field of vision,and the search for convenient interaction has become a new research topic.In order to get rid of the traditional machine-centered interaction mode and achieve the purpose of improving human-computer interaction,a gesture recognition control system based on wheeled inverted pendulum robot is designed,which realizes the motion state of gesture control robot.The wheeled inverted pendulum robot used in this system is based on the Beaglebone Black development board.By extending the photoelectric encoder,attitude detection module and other sensors,combined with the PID control,the basic motion functions of the forwarding and steering of the wheel inverted pendulum robot are completed.Firstly,in order to obtain the gesture target,a method of combining skin color model detection and moving target detection is proposed to complete the gesture segmentation work.Motion detection is performed on the acquired gesture image,the foreground is separated from the background,a motion region is obtained,and the gesture target position is further determined by performing skin color detection on the motion region.Secondly,according to different types of gestures,different gesture feature extraction methods are adopted respectively.The extracted HOG feature and LBP operator from the image are serially fused into a static gesture feature vector.Unlike the static gesture,the gesture centroid position is obtained by performing Kalman filter for tracking on the dynamic gesture.In the case of ignoring the shape of the gesture,the centroid position information is used to calculate the direction angle,and then the dynamic gesture feature vector is obtained.Thirdly,the identification of nine custom static gestures which is based on the support vector machine classification algorithm is completed.Depending on whether the dynamic gesture changing the hand shape or not during the motion,the matching of the six dynamic gesture trajectories is completed,which is based on the dynamic time warping algorithm.Finally,through the semantic recognition of the custom gesture,the recognition result is converted into a control command and sent it to the robot by the WiFi communication system,and then the robot completes the motion steering in the light of the instruction.The experiment result proves that the system of the hand gesture recognition motion control system of the inverted pendulum robot meets to the design requirements.The effectiveness of the system is fully demonstrated,and the method has versatility and the gesture action can be extended.