A Human Body Based on Sift-Neural Network Algorithm Attitude Recognition Method

The selection of highly distinctive features and appropriate recognition strategies is the key to human gesture recognition technology. This paper discusses the method of human body pose recognition based on affine invariant geometric feature sift operator. The calculation complexity of SIFT operator is high, and different body posture and image blur will increase the difficulty of feature recognition. In order to overcome the above shortcomings, an improved sift algorithm is proposed, which introduces image classification and block processing into subregions of interest for description. Finally, BP neural network is used to establish a human gesture recognition model and improve the accuracy of human gesture recognition. Experimental results show that this method can not only effectively reduce the influence of posture difference and image blur on the reduction of posture recognition rate of standing, bending and falling, but also significantly improve the calculation speed and gesture recognition efficiency of sift operator.