基于特征提取极限学习的卧姿压力图像识别  

Image recognition of lying posture pressure based on feature extraction extreme learning

作　　者：袁先启 蒋永翔[1] 夏红超 YUAN Xianqi;JIANG Yongxiang;XIA Hongchao(Institute of Robotics and Intelligent Equipment,Tianjin University of Technology and Education,Tianjin 300222,China)

机构地区：[1]天津职业技术师范大学机器人及智能装备研究院,天津300222

出　　处：《天津职业技术师范大学学报》2022年第2期58-62,共5页Journal of Tianjin University of Technology and Education

基　　金：天津市科技支撑重点项目(18YFZCSF00600)。

摘　　要：卧姿是识别评判睡眠质量和预防突发疾病的指标之一,文章基于极限学习算法(ELM)对卧姿压力图像进行了研究。采用阵列式柔性压力传感器获取背部压力图像,通过对图像预处理,完成了图像几何特征值、能量特征值、颜色特征值的提取;引入ELM算法对16种不同特征值进行训练预测。结果表明:在1 280组特征值中,将1 120组作为训练数据,160组作为测试数据,当隐藏节点为80时,卧姿识别的正确率为98.75%。This paper explores the recumbent pressure images based on the extreme learning algorithm(ELM) for lying posture is one of the indicators to identify and evaluate the sleep quality and to prevent sudden diseases. The back pressure image is acquired by using an array of flexible pressure sensors,and the extraction of geometric feature values,energy feature values,and color feature values of the image were completed by pre-processing the image. The ELM algorithm was introduced to train the prediction of 16 different feature values. The results show that 1,120 sets among a total of1,280 sets of feature values are used as training data and 160 sets are used as test data,and the correct rate of lying posture recognition is 98.75% when the hidden node is 80.

关 键 词：卧姿识别 极限学习算法(ELM) 图像特征值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]