Research Of Human Activity Recognition Method And Its System Implementation Based On Recurrent Neural Network

At present,most human activities recognition research is clustered on video data and the activity recognition method based on video data that the recognition accuracy is not high due to the complexity of video image data is easy to violate personal privacy and.With the rapid development of Internet of Things technology,the use of wearing equipment to recognize human activities has gradually become a new method of human activities recognition attracting a large number of experts and scholars of the eyes.Researchers try to use a number of machine learning methods,such as random forest,support vector machine and other shallow learning methods.In the laboratory environment,human activities recognition methods based shallow learning are good performance,but there is still some distance from the practical application.Because human activities belongs to a continuous time series data and the recursive neural network in the time series data show the magic effect.At the moment a large number of practical applications based on the recursive neural network have achieved good results.This paper uses recursion neural network model based on the deep learning to solve human activities recognition.The specific research work and innovation of this paper include:1.Based on the characteristics of human behavior,this paper presents a hierarchical identification of the behavior of the method.2.This paper defines atomic activity as a short cycle of activities and outbound traffic activities,including falls,downstairs,upstairs,elevator down,lift up,run,sit,stand,take the subway,take the bus,take the car,lying 13 kinds of activities.The recursive neural network model for atomic activity recognition is used to construct the recognition algorithm.At the same time,a large number of experiments have been carried out and good results have been obtained in the laboratory environment and the practical application environment.3.This article defines complex activities as long-term life activities,including but not limited to cooking,mopping,sweeping,eating,watching TV and so on.In this paper,a complex activity recognition algorithm based on dynamic time warping is designed with reference to the idea of pattern matching in time series analysis.The algorithm can recognize three kinds of complex activities such as sleeping,eating and exercising in the laboratory environment,and has the lower average index error(the evaluation index of the complex activity recognition algorithm,see chapter 4.2 for details).4.This paper implements a kind of intelligent system for the recognition of prototyping systems,such as data acquisition,data analysis,data display and data management,and the wisdom of the research results.