【摘要】 为了准确区分各种肺音信号,获得更理想的肺音识别效果,提出了一种基于语谱图的改进型LBP肺音识别方法.首先通过短时傅里叶变换将肺音信号转化为灰度语谱图;其次利用改进后LBP算法计算语谱图的局部纹理关系,将局部二值模式特征进行级联构成特征向量;最后利用支持向量机对正常肺音和三类异常肺音信号进行识别分类.结果表明,该方法对不同肺音信号的识别率可达92.59%,为肺部疾病的医疗诊断提供了新的思路.更多还原

【Abstract】 In order to accurately distinguish various lung sound signals and obtain better lung sound recognition effects,an improved LBP lung sound recognition method based on spectrogram was proposed.Firstly,the lung sound signal is converted into a gray spectral map by a short-time fourier transform,and then the local texture relation of the spectral map is calculated by using an improved LBP algorithm.The local binary pattern features are cascaded to form a feature vector,and finally a support vector is used.The machine classifies and recognizes normal lung sounds and three types of abnormal lung sound signals.The results showed that the recognition rate of this method was 92.59%for different lung sound signals,which provided a new idea for medical diagnosis of lung diseases.更多还原