PTB诊断心电数据库中的原始ECG信号的特征提取

论文编号:DZXX370 论文字数:10697,页数:47

摘要

 传统特征提取的方法是从类间平均可分性最优化观点出发的，企图对其所涉及的所有类别进行分类。然而，在许多情况下，并非所有类别都有着同样的重要性和实际意义，而这种无差别的运算极大地增加了运算时间，为此，我们希望能够提出一种全新的算法，在特征提取时在给予最重要的类别一定的优先分类权的同时其它各类也尽可能得到分类，其主要思想是: 将人的主观意图融合到特征提取中, 提取有利于具有优先权类别的分类特征，我们将其命名为主观引导特征提取法。结果表明，与传统线性鉴别分析法 （LDA）相比较，该方法可获得更加有效的分类特征。

关键词：信息，特征提取，可分性，优先权分类

ABSTRACT
 
 Conventional feature extraction methods extract features from the viewpoint of the best average separability among the classes. Lots of studies on feature extraction for classification attempt to classify all the classes included. However sometimes it is more important and significant to recognize certain specific classes (prior classes) rather than to discriminate all the classes. Such non-discriminatory computing greatly increase the computing time, to that end, we hope to propose a new algorithm, The extracted features can discriminate the prior class from others and separate each of the classes as much as possible at the same time. The subjective introduction based method was proposed, which the subjective significance was introduced into the feature extraction for the classification with priority in the paper, We will rename its subjective proposed feature extraction method. The experimental results show that the method can obtain more effective features compared to conventional linear discriminant analysis (LDA) methods.

Keywords: Informaiton, feature extraction, separability, classification with priority
                                                                    

目录

摘要 I
ABSTRACT II
目录 III
第1章 绪论 1
1.1 课题的背景和意义 1
1.2 国内外研究现状 1
1.2.1 线性鉴别分析法（LDA）及其改良算法 2
1.2.2 主元分析法（PCA）及其改良算法 2
1.2.3 特征提取算法的发展趋势 2
1.3 课题的方法与结论 3
1.4 MATLAB介绍 3
第2章 设计方法 4
2.1 ECG信号 4
2.2 特征提取 5
2.3 设计流图 5
2.4 实验数据说明 6
2.5 K-L变换 6
2.5.1 K-L展开式 7
2.5.2 K-L展开式的性质 8
2.5.3 K-L坐标系的产生矩阵 9
2.5.4奇异值分解(SVD)定理 10
2.6 白化 11
2.7 特征投影相量的形成 11
2.8 线性鉴别分析法（LDA） 12
第3章 相关过程分析 15
3.1 ECG信号提取 15
3.2 K-L变换 16
3.3 白化 17
3.4 特征投影相量的形成 18
3.5 分类结果分析 22
第4章 对比实验及结果分析 23
4.1 主观引导算法的特征提取结果 23
4.2 线性鉴别分析法（LDA）的特征提取结果 24
4.3 实验结果分析 26
第5章 结论 27
致谢 28
参考文献 29
附件1 程序代码 30