Study And Implementation Of Personal Fitness Management Base On Human Activity Recogniton On Mobile Platform

Currently, smartphone has deeply integrated into people’s lives. And current smartphones have made great progress in computing and storage capacity, connection mode, and human-computer interaction. Especially current smartphones have a lot of built-in sensors. These conditions provide a good opportunity to study the behavior of smartphone users. We can analyze the user behavior through mining the sensors data built in smartphone and provide more personalized and intelligent service to user.In this thesis, we first study of personalized human activity recognition(HAR) classifier model. Then we design and implement a HAR core library which can recognize user’s activities including walking, jogging, cycling, sitting and standing on android. Base on this library, we developed a fitness management mobile application which can evaluate user’s fitness. The main content of this thesis including:Firstly, we study the key technologies of HAR on android which includes the HAR raw dataset collect, the preprocessing of raw dataset, and the training and performance test of classifier. Then we study the personalized HAR classifier model which includes comparing universal classifier model with personalized classifier model and obtaining a learning curve through experiment.Secondly, Base on the study of key technologies of HAR, we design and implement a HAR core library which includes core feature of HAR on android. Base on this library, developers can easy to add HAR function to their application.Thirdly, design and implement a mobile application of fitness management. We first design the architecture of application. Base on the HAR library, application can recognize user’s activities on background. Then application can evaluate the user’s fitness status through these recognition results. Application also implement the personalization of HAR classifier which base on the study before.This thesis has completed the content above. Study and implement of key technologies of human activity recognition and application of personal fitness management on smartphone. After test, the function and perform of system meet the requirements. This thesis can be reference for researcher and developer who need to implement HAR on android.