基于VMD分解的呼吸音频谱特征在COPD识别中的应用

Application of VMD⁃based respiratory sound spectrum features in COPD recognition

慢性阻塞性肺病(COPD)早期诊断技术对疾病的治疗具有重要的作用,文中提出基于呼吸音最大瞬时频率(f_(Emax))特征的COPD疾病识别算法。算法选取指定k值的变分模态分解(VMD)将呼吸音分解,提取本征模态函数(IMF)分量,再对该IMF分量做Hilbert变换(HT)计算边际谱,得到最大瞬时频率f_(Emax)、瞬时频率幅值及特征能量等频谱特征参数。最后,利用主成分和线性判别分析对健康人群和COPD患者的呼吸音频谱特征参数进行研究,分析COPD与上述频谱特征参数的相关性。实验结果表明,上述参数中,f_(Emax)与COPD疾病显著相关,能够较好地区分健康人群和COPD患者,识别精度高达92.62%,可以作为COPD疾病早期筛查的一种辅助手段,具有较好的实际应用前景。

Early diagnosis technology of the chronic obstructive pulmonary disease(COPD)plays an important role in the treatment of the disease.In view of this,a COPD recognition algorithm based on the maximum instantaneous frequency(fEmax)feature of respiratory sounds is proposed.In the algorithm,the respiratory sound is decomposed by the variational mode decomposition(VMD)with the specified k value,the intrinsic mode function(IMF)component is extracted,and then,the IMF component is subjected to Hilbert transform(HT)for the calculation of marginal spectrum,so as to obtain the spectrum characteristic parameters of the maximum instantaneous frequency fEmax,instantaneous frequency amplitude and characteristic energy.The characteristic parameters of the respiratory sound spectrum of the healthy people and the patients with COPD were studied by principal component analysis(PCA)and linear discriminant analysis(LDA).The correlation between COPD and the above spectrum characteristic parameters was analyzed.The experimental results show that,among the above parameters,f_(Emax)is significantly correlated with COPD,which can better distinguish the healthy people from the COPD patients,and its recognition accuracy can reach 92.62%,so it can be used as an auxiliary means of early screening of COPD.Therefore,the proposed recognition method has a good practical application prospect.