Design And Implementation Of A Kinect-based Push-up Movement Evaluation System

The aim of assessing movements is to automatically analyze and evaluate the body movements of humans,which is often applied in sports and fitness testing in universities and military contexts.Currently,the push-up movement is evaluated manually as part of physical fitness tests.This approach is inefficient and subjective,leading to significant errors in the assessment results.In this paper,the key joint points describing the push-up movement are selected according to the National Physical Fitness Standard,and by comparing and analysing two human skeletal joint point detection methods,Open Pose and Kinect,selected Kinect-based joint point detection method,detect the joint point data of push up motion and calculate the 8 joint angles of push up as motion characteristics.Ultimately,this paper presents a novel approach to assessing push-up movements using the dynamic time regularization algorithm.Additionally,the authors have developed and executed a push-up movement evaluation system that is based on Kinect technology.To summarize,the main research work in this paper includes:1.This paper compares and analyzes the Open Pose and Kinect methods for detecting limb joints in the context of evaluating push-ups,considering the latter’s advantages of higher real-time detection accuracy and speed,more detected joints,and lower machine configuration.Based on these factors,the Kinect-based detection method is selected as the push-up joint detection method.To address issues of hardware performance and external environmental interference,the article proposes using a Kinect depth camera to extract skeletal joint coordinate data and applying a smoothing filter algorithm to process joint coordinate data.Based on the push-up joint coordinate data,eight joint angles are calculated as push-up motion features,and the invariance of these features is verified..2.This paper deals with the challenge posed by the variable length of time-action sequence curves resulting from the diverse movement rates of the human body.It uses the Dynamic Time Warping algorithm(DTW)to match the feature curves of push-up actions,and combines this algorithm with an improved strategy to design a hybrid matching algorithm(Fixed DTW,FDTW)for calculating the similarity between user action sequences and standard action sequences.The paper then conducts corresponding push-up action evaluation experiments,and the results show that FDTW algorithm outperforms DTW algorithm in terms of time complexity,with a performance improvement of 38.47%.Finally,the paper summarizes a push-up evaluation formula for quantifying push-up actions.3.This paper presents the design and implementation of a push-up movement evaluation system based on Kinect technology.We will divide the system into different levels.Each layer has its own set of responsibilities and functions.They are: business layer,service layer,data layer and algorithm layer.The business layer combines the intelligent push-up evaluation requirements of universities and designs the business model;the service layer is responsible for designing the web page and Electron application to facilitate interaction with users,while the data layer employs a My SQL database and the file system Fast DFS to manage and save the data;to enable interaction with the system,the algorithm layer utilized the Flask framework in Python to encapsulate the FDTW algorithm within an HTTP address interface,which was subsequently integrated into the Electron application.The system underwent functional and performance testing in the end,which showed that it achieved the design objectives in terms of function and performance,and also verified the feasibility of the evaluation method in the actual evaluation scenario presented in this paper.