基于LDA和小波分解的肺音特征提取方法

发布时间：2018-08-14 14:41

【摘要】：针对用小波分解提取肺音特征后特征向量维数较高的问题,提出了一种结合线性判别分析和小波分解的肺音特征提取方法。在该方法中,首先对肺音信号进行小波分解,然后将小波分解得到的小波系数转化成小波能量特征向量,接着使用线性判别分析法对该特征向量进行降维处理,得到新的低维特征向量,最后用SVM对低维特征进行识别。在实验中,选取了三种肺音信号:正常肺音、爆裂音、哮鸣音,用所提出的方法将8维的小波能量特征降为2维特征,在2维特征上进行了分类识别,并和降维之前的结果进行了比较,实验结果表明利用线性判别分析对小波能量特征降维后极大地提高了识别精度。同时,和其他几种典型的肺音特征提取方法进行了比较,实验结果表明结合线性判别分析与小波分解的特征提取方法得到了更高的识别精度。
[Abstract]:In order to solve the problem of higher dimension of feature vector after extraction of lung sounds by wavelet decomposition, a method of feature extraction of lung sounds based on linear discriminant analysis and wavelet decomposition is proposed. In this method, the lung sound signal is first decomposed by wavelet, then the wavelet coefficients obtained by wavelet decomposition are transformed into wavelet energy eigenvector, and then the dimension reduction of the eigenvector is processed by linear discriminant analysis (LDA). A new low dimensional feature vector is obtained. Finally, SVM is used to identify the low dimensional feature. In the experiment, three kinds of lung sound number are selected: normal lung sound, burst sound, wheezing sound, and the proposed method is used to reduce the 8-dimensional wavelet energy feature to 2-D feature, and to classify and recognize the 2-D feature. Compared with the results before dimension reduction, the experimental results show that the recognition accuracy is greatly improved by using linear discriminant analysis (LDA) to reduce the dimension of wavelet energy features. At the same time, compared with other typical methods, the experimental results show that the feature extraction method combined with linear discriminant analysis and wavelet decomposition has higher recognition accuracy.
【作者单位】：河北工业大学计算机科学与软件学院;河北省大数据计算重点实验室;
【基金】：天津市应用基础与前沿技术研究计划重点项目(No.14JCZDJC31600) 河北省自然科学基金专项(No.F2016202144) 河北省高等学校科学技术研究重点项目支持(No.ZD2014030)
【分类号】：TN912.3

【相似文献】