Recognition Of Human Action In The Video Based On The Rate Of Change In Range Of Motion

Recognition of human action is a hot research field of machine vision today, It has great value,The main contents of the study are as follows: detection torso area, detection of characteristic parameters,action split and gesture recognition. Segmentation method of human action based on rate of change of amplitude for crunches in this paper, The continuous crunches is divided into different attitude by this method. Check the correctness of the division method on the other hand. Dividing unit operation is used to split video into crunches unit. Gesture segmentation is used to distinguish critical attitude and non-critical attitude. Critical attitude can be used to judge crunches qualified or unqualified. Experiments show that this method is feasible, It has the following advantages: Less operation Real-time high and so on.The main contents are as follows:1, In this paper, for crunches exercise, we proposed a gesture segmentation method based on the characteristics of the change rate of action. First, extracting frame from the video. Then extraction contour by background subtraction. Next to describe the body’s movements and gestures by aspect ratio.Finally verify the correctness based on experiment.2, This paper presents an algorithm to eliminate errors, this algorithm is based on the magnitude of a vector, and it is used to eliminate the errors caused by the abnormal data in the data sequence. The basic principle is to build a set of vectors using several adjacent data.3, Experimental verification results are as follows: Threshold delineation and recognition rate is related. On the other hand, the key gesture recognition method is proved to be effective by visual method, it means that it is possible to reduces workload by this method.