基于移动最小二乘的传感器数据重构.pdfVIP

基于移动最小二乘的传感器数据重构 　宋绍民,等 基于移动最小二乘的传感器数据重构 D a ta Recon struction for Nanosen sor by U sing M ov ing L ea st Square M ethod 1 2 1 宋绍民 　王 　敏 　胡新晚 ( 1 2 ) 湖南工学院电气与信息工程系 ,湖南 衡阳　42 1002;湖南大学电气与信息工程学院 ,湖南 长沙　4 10082 摘 　要 : 纳米气体传感器的灵敏度易受温度变化的影响。为了有效改善纳米传感器的温度特性 ,方便确定其最佳工作条件 ,利用移 动最小二乘法对其灵敏度 - 温度数据进行拟合 ,并建立相应的近似模型 。移动最小二乘法通过紧支性加权的局部近似 ,使所建模型 较常规最小二乘模型更精确 、更光滑 。实验证实了基于移动最小二乘拟合的近似性能。该方法在传感器数据精确重构领域具有 良好 的应用前景 。 关键词 : 传感器 　数据重构　温度特性 　移动最小二乘法 　近似技术 　曲线拟合 中图分类号 : TP2 12. 1　　　　文献标志码 : A A bstract: Generally, the sen sitivity of nanoga ssen sors is easily affected by variation of the temp erature. In order to effectively imp rove the temp erature characteristic s of the sen sor, and be ea sy to determ ine its op tim al op eration condition s, by adop ting moving least square m ethod, the sen sitivitytemp erature data are fitted, and the app roxim ate model is e stab lished. Through comp act support weighted local app roxim ation , the model estab lished w ith moving least square (ML S) is more accurate and smooth than the model estab lished w ith conventional lea st square m ethod. The exp erim ents p rove the app roxim ate p erform ance of the fitting ba sed on ML S. Th is m ethod po ssesses excellent p ro sp ects in p recise recon struction area of sen sor data. Keywords: Sensor　Data recon struction　Temp erature characteristic　Moving least square method　Technology of app roximation　Curve fitting 0　引言 1　ML S拟合 纳米气体传感器具有响应速度快 、稳定性好 、