An Improved Hu-moment Algorithm in Gesture Recognition Based on Kinect Sensor

摘要

In order to solve the problem of variability of gesture recognition, The Kinect sensor and the improved Hu-moment is used in this study. Depth image, after image restoration, its feature data is extracted using depth threshold is obtained based on Kinect. Then the scale factor is added to the Hu-moment, so that the invariant moments contain more detail characteristics of gesture and have relationship with geometry but not scaling. For illustration, a dress fitting system example is utilized to interpret the algorithm. Empirical results show that the gesture recognition method is performed correctly in illumination and complex background. So the method of gesture recognition has invariance in image rotation, scaling and translation and has robustness to background interference.