Research,Design And Implementation Of Human Activity Recognition Information Processing System Based On Lstm Recurrent Neural Network Algorithm

Recently,with the popularity of sports wearable smart devices,the acquisition of data on human movement has not become difficult,and has spawned a series of running fitness software,which has led the wave of running for the whole people and greatly promoted the development of the sports industry.However,a large amount of activity data has not been deeply mined,which has caused huge waste of its value.In order to make the data collected by sports smart devices better serve sports enthusiasts and thus more effectively improve the information degree of sports industry,this paper does research in the design and implementation of human activity recognition information processing system as the main research content.By combing the previous research results of human activity recognition system about sports wearable smart devices,this paper summarizes a three-layer human activity recognition information processing system architecture.In the data calculation layer,different from the traditional machine learning classification algorithm,a classifier based on LSTM recurrent neural network algorithm is proposed.As a kind of deep learning,the LSTM is good at processing long and long intervals and has correlation.The data and the characteristics of the features in the automatic learning data,in order to solve the defects in the traditional identification method that need to manually extract the activity features from the data,so that the entire system is streamlined and more efficient.This research solves the following problems:(1)Based on the research of existing human activity recognition information processing system,a three-layer human activity recognition information processing system is designed,including data acquisition layer,data calculation layer and data application layer.The data acquisition layer collects and preprocesses the data;the data calculation layer is the core layer of the system,responsible for data storage,segmentation,feature extraction and activity recognition classification;the data application layer is used to visually display the recognition results and activity data.The three-layer human activity recognition information processing system is designed and implemented by software engineering,rapid prototyping and Python computer language.The system can input,store,recognize and display human activity information on the web.(2)In order to overcome the shortcomings of traditional human activity recognition information processing system,such as manually selecting human movement characters,cumbersome recognition process and low recognition efficiency,LSTM recurrent neural network algorithm is used as classifier in data calculation layer in human activity recognition information processing system,which can automatically extract multi-dimensional features directly from basic features and avoid manual work.The time domain and frequency domain features of time series data are extracted,and the loss of data information caused by dimensionality reduction is avoided.The LSTM recurrent neural network model with daily activityrecognition is trained by using the open data set PAMAP2.The performance evaluation results of the algorithm show that the algorithm has excellent ability of daily activity recognition and mining.