Research On Human Activity Recognition Classification Method With Sparse Representation

In recent years,action recognition technology has gradually become a hot spot for scholars in smart home.A good feature extraction method plays an important role in the action recognition.In the current studies of recognition recognition,the main problems are:the sequential human behavior,in the same behaviors which people exhibit,the order between different time sequences and the potential link between human behaviors;the sensitivity of similar human actions,because of theirs' similar fragment between many different behaviors;the complexity of human behavior,they have different expressions for the same kind of behavior.In this paper,the basic purpose of sparse representation action recognition is to enhance the expression ability of the model to the feature vectors,so that the frame can be identified more accurately.In view of the current main feature extraction methods,most of them lack the miningl of the potential rules,which leads to the weak expression ability of feature extraction.Moreover,these features involve too many repeatability sensor event information,which makes the feature redundant.In addition,the generality of these models are weak for different smart home environments.In order to solve the problems of behaviors' time-series and complexity,this paper first proposes a action recognition feature extraction method based,which mining the rules between different behaviors using rough sets to dig out more information of behavior and improve the expression of behavior ability.Secondly,we propose a feature extraction action recognition method,this method is applied to continuous data and make it discretized,which makes the continuous data more practical.Finally,a action recognition method based on sparse representation dictionary training is presented in this paper,aiming at improving the sensitivity between different actions and making it more accurate for selection the corresponding dictionary.Through the simulation experiments for these three behavior recognition methods,which obtained good research results,and the effectiveness of the algorithm is confirmed.