基于卷积神经网络的大学男生体脂率预测  

Prediction of body fat percentage of male undergraduates based on convolutional neural network

作　　者：郝霖霖 赵喜迎[2] HAO Linlin;ZHAO Xiying(Fudan University,Shanghai 200433,China;Nanjing Xiaozhuang College,Nanjing 211171,Jiangsu,China)

机构地区：[1]复旦大学,上海200433 [2]南京晓庄学院,江苏南京211171

出　　处：《辽宁体育科技》2022年第5期76-84,共9页Liaoning Sport Science and Technology

基　　金：教育部人文社会科学研究项目(项目编号:18YJC890007);教育部人文社会科学研究青年基金项目(项目编号:19YJC890062)。

摘　　要：建立混合数据输入的卷积神经网络回归模型,利用图像和BMI数据预测体脂率,构建混合数据输入的CNN回归模型,测试模型的精确度。解决以往体脂率预测模型无法同时适用于健身群体和非健身群体的问题。结果显示:构建的混合数据输入的CNN回归模型预测精确度达到94.62%,绝对误差为1.57%,要优于单纯用BMI作为自变量构建的线性回归模型或ANN回归模型。可见,建立的混合数据输入的CNN回归模型,预测体脂率的精确度较高,能同时适用于大学男生健身群体和非健身群体,具有较高的便捷性、经济性和实用性。In order to solve the previous body fat percentage prediction model problem that cannot be applied to fitness groups and non-fitness groups at the same time,this paper establishes a convolution neural network regression model with mixed data,and predict the body fat percentage by using image and BMI data.The results show the prediction accuracy of CNN regression model with mixed data was 94.62%and the absolute error was 1.57%,which was better than linear regression model with BMI as independent variable or ANN regression model.The CNN regression model with mixed data has high accuracy in predicting body fat percentage,can be applied to both fitness group and non-fitness group,and has convenience,economy and practicability.

关 键 词：体脂百分比 卷积神经网络 身体质量指数 大学男生 

分 类 号：G804.49[文化科学—运动人体科学]