文章摘要

文章信息/Info

Title:: Measurement anddata processing of human neck and waist physiological curve based on time-of-flight

作者:: 孙琦1; 严荣国1△; 江容安2; 刘亚萍1; 1.上海理工大学医疗器械与食品学院，上海 200093；2.上海市杨浦区中心医院康复医学科，上海 200082

Author(s):: SUN Qi; YAN Rongguo; JIANG Rongan; LIU Yaping; 1.School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China;2.Department of Rehabilitation Medicine, Yangpu District Central Hospital, Shanghai 200082

Keywords:: Curve of neck and waist; Time-of-flight laser ranging; Single factor analysis; Regression analysis; K-nearest neighbor algorithm

摘要:: 设计一款人体颈腰部生理曲线测量装置并对采集到的生理曲线数据进行分析。数据采集装置主要是基于飞行时间（time-of-flight，TOF）测距原理，分为曲线测量装置、数据采集装置、计算机曲线拟合等三个模块。数据分析部分包括基于SPSS 22平台的单因素分析、回归分析，基于Python的K近邻算法分类识别。对受试者的身高、体重、年龄和矢状面垂直距离（SVA）做Pearson相关性分析，对性别和SVA做t检验。对身高、体重、年龄和矢状面垂直距离（SVA）做回归分析。选身高、体重、年龄和矢状面垂直距离（SVA）作为参考指标，将年龄作为标记对象（青少年、成人、老年人）对曲线做K近邻算法分类处理。身高、体重、年龄和矢状面垂直距离（SVA）具有相关性（P<0.05）；性别和SVA值不相关（P>0.05）。用K近邻算法可以较好对曲线做分类，识别准确率为74.93%。

Abstract:: To design a device for measuring the physiological curve of human neck,waist and analyze the collected physiological curve data. Data acquisition device was mainly based on the principle of time-of-flight(TOF) flight time ranging. It was divided into three modules: curve measurement device, data acquisition device and computer curve fitting module. The data analysis part was divided into single factor analysis based on SPSS 22 platform, regression analysis,K-nearest neighbor algorithm(KNN) classification and recognition based on python. Pearson correlation analysis was performed for height, weight, age and SVA, and t test for gender and SVA. Regression analysis was made on height, weight, age and sagittal vertical axis (SVA). Height, weight, age and SVA were selected as reference indexes. Age was used as marker objects (adolescents, adults and the elderly) to classify the curve by KNN.Height、weight、age and SVA were correlated (P<0.05), while gender and SVA were not correlated (P>0.05). KNN can be used to classify curves, and the recognition accuracy is 74.93%.