Preprocessing And Similarity Detection Of Sports Bracelet Heart Rate Series

It is reported that the students' physique condition shows a decline trend in recent years.Consequently many universities have enforced early morning fitness running.We claim that a better solution for this issue is to use sports bracelets to supervise fitness running.One of the key technique challenge is to prevent “pinch runner” who wears multiple bracelets to help others meet the extracurricular exercise requirement.In order to improve the quality of sports bracelet heart rate data,we first propose a sliding-window-based regression model to calibrate them.Then we propose a SVM–based fake running detection solution to prevent pinch runner.The main contributions of this thesis are as follows:1.The thesis proposes a sliding-window-based linear regression model. Samples are generated according to the relationship between sport bracelet series and heart rate band series.This method can calibrate all sports bracelet heart rate series effectively.2.The thesis proposes a SVM–based fake running detection solution.The solution extracts distance related and statistical features from all time-overlapped series pairs to generate samples.A linear kernel SVM is trained from these samples and providing precise and high recall pinch runner detection.This method successfully detects pinch runner during management of student fitness running.The effectiveness of the proposed method is verified by experiments.