Arm Swing Recognition in Sprint Training Based on Multi-scale Detail Enhancement
There are problems in the method of arm swing action recognition that the image features are directly ex-tracted,resulting in the low fit of the arm action recognition result and the actual arm movement trajectory.The method of the arm swing action recognition in sprint training based on multi-scale detail enhancement is proposed.The three steps of image multi-scale decomposition,image texture detail enhancement and image information fusion are used to highlight the action characteristics of the swing arm and identify the arm swing in sprint training.The experiment results show that the deviation of trajectory identified by this method and the actual trajectory of the swing arm is small,which has a better recognition effect of the arm swing.
swing arm actionaction recognitionsprint trainingmultiscale detail enhancementLSTM network
NETL
NSTL
万方数据
