Abstract:
In order to explore the contribution rate of each feature in the initial feature set for identification,under the guidance of recognition accuracy rate,this paper uses stepwise discriminant analysis method to acquire the contribution rate and sorting of each feature,and select a key feature subset used for identification.Experiments are carried on Physikalisch-technische Bundesanstalt(PTB)and lab collected data sets which select 9 and 17 key features from two experiment data sets respectively.Experimental results show that these features have 66.7% coincidence degree,while the fiducial points of the two sets depending on have 63.6% coincidence degree,and initial feature set of Electrocardiogram(ECG) contains common features to distinguish individuals to a certain extent.Meanwhile,using the two key feature subsets,the recognition rate of two data sets is 99.7% and 94.8% respectively.
Key words:
Electrocardiogram(ECG),
identification,
fiducial point,
feature selection,
stepwise discriminant analysis,
initial feature set,
key feature subset
摘要： 研究初始特征集合中各个特征对身份识别的贡献率,依据身份识别准确率,采用逐步判别分析法,确定识别系统中各个特征的贡献率和排序,挑选出可用于身份识别的关键特征子集。利用PTB心电数据库和实验室自采的心电数据进行实验,在2个数据集中分别选出9个和17个关键特征,结果表明,特征重合度达到66.7%,特征所依赖的波形基准点重合度达到63.6%,选取的心电波形初始特征集合包含了可区分个体差异的公共特征,2个数据集利用各自获得的关键特征子集进行身份识别测试时的准确率分别达到99.7%和94.8%。
关键词:
心电信号,
身份识别,
基准点,
特征选择,
逐步判别分析,
初始特征集合,
关键特征子集
CLC Number: