时间序列降维及机场噪声中的机型识别  

Time series dimensionality reduction and aircraft model recognition in airport-noise

作　　者：王寅同[1] 王建东[1] 陈海燕[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,南京210016

出　　处：《吉林大学学报（工学版）》2016年第4期1202-1208,共7页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金重点项目(61139002);中央高校科研业务费专项资金项目(NS2015091);江苏省博士后科研资助计划项目(1301013A)

摘　　要：为了提高非完整标记的高维机场噪声数据的处理速度和效率,研究了时间序列降维及机场噪声中的机型识别问题。首先采用概率类和不相关判别的半监督局部Fisher方法(SLFisher)得到降维转换矩阵,再将时间序列数据由高维空间映射到低维空间,最后在低维数据上进行k最近邻分类(kNN)。在国内某机场的实测噪声数据上的实验结果表明,SLFisher降维后机场噪声事件数据的机型识别效果取得显著提升。In order to enhance the efficiency and effectiveness of the processing of high dimensional airport-noise data with incomplete labels,the time series dimensionality reduction and aircraft model recognition in airport-noise are investigated.First,we use semi-supervised local fisher method,which is based on probability class and uncorrelated discriminant（SLFisher）,to obtain the transformation matrix of dimensionality reduction.Next,the high dimensional time series data are mapped to the low-dimensional space.Finally,we apply the k Nearest Neighbor（kNN）classifier to classify the obtained low-dimensional data.Experimental results on measured airport-noise data demonstrate that the performance of aircraft model recognition is remarkably improved after the dimensionality reduction achieved using SLFisher.

关 键 词：人工智能 机场噪声 时间序列 降维 机型识别 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]