喘鸣音的声谱图熵特征分析及检测
Wheeze detection method based on spectrogram entropy analysis
-
摘要: 提出了一种改进的基于肺音信号的声谱图熵特征分析的客观喘鸣音检测方法。喘鸣音的功率明显高于正常肺音,因此喘鸣音声谱图的功率分布沿频率轴方向具有明显的聚集特性,该特性可以通过熵值反映。本算法首先对肺音信号进行时频变换得到时频幅度谱信号,然后去除基本呼吸音,进而计算其熵特性曲线并提取熵特性曲线的相应特征.最后,通过支持向量机(support vector machine,SVM)训练分类器,实现了喘鸣音的有效检测。该方法通过预处理使熵特性曲线的特征差异更加明显,且通过SVM分类器进行检测,解决了原方法检测存在检测模糊区域的问题。实验结果表明,该算法在两组测试集的检测准确率分别为97.1%和95.7%,检测率较高,具有良好的应用前景。Abstract: A wheeze detecting method based on spectrogram entropy analysis was introduced.The power of wheeze is much higher than that of normal lungsounds,so the wheeze has clustering characteristics in terms of power distribution along frequency axis,which can be reflected by the entropy.This algorithm mainly comprises three steps which are the short-time Fourier transform (STFT) decomposition and detrend,the calculation of the spectrograms entropy and extraction of the features,and wheeze detection based on support vector machine (SVM).The step of detrending facilitates the difference of the features between wheeze and normal lung sounds.Moreover,compared with the original method,there is no uncertain decision any more.The result of the two testing experiments shows that the accuracy (AC) are 97.1%and 95.7%,respectively,which proves that this method could be an efficient way to detect wheeze.
下载: