Solitary oldies abnormal action recognition based on MEI

This paper presents a new method for identifying the abnormal action of the solitary oldies which is based on video sequence. First, we use the background subtraction and morphological filtering technology to extract the moving human contour. Then, we extract the motion energy image (MEI) of the moving body target, which is followed by extracting the Hu moments feature of human motion energy image extracted. At last, we classify and identify the abnormal action by using Bayesian classifier. Experiments demonstrate that the proposed recognition method is simple and practical. It achieves the correct recognition rate of daily behavior more than 92%. This method can also well identify the falling action, its recognition result is more ideal, and has some great practical value.