Crane Guidance Gesture Tracking and Recognition With Nonlinear Estimation and Fuzzy Logic

This paper presents a novel human arm gesture tracking and recognition technique based on fuzzy logic and nonlinear Kalman filtering with applications in crane guidance. Kinect visual sensor and MYO armband sensor are jointly utilized to perform data fusion in providing more accurate and reliable information on Euler angles, angular velocity, linear acceleration and electromyography data in real-time. Dynamic equations for arm gesture movement are formulated with Newton-Euler equations based on Denavit-Hartenberg parameters. Nonlinear Kalman filtering techniques, including the extended Kalman filter and the unscented Kalman filter, are applied to perform reliable sensor fusion, and their tracking accuracies are compared. A Sugeno-type fuzzy inference system is proposed for arm gestures recognition. Hardware experiments have shown the efficacy of proposed method for crane guidance applications.