Research On The Diagnosis Method Of Lung Diseases Based On The Analysis Of Breath Sound Characteristics

With the increase of age and the aging of body functions,people are at greater risk of suffering from lung diseases.Once the lung organs have health problems,it will not only be difficult to treat but also directly threaten the safety of life.Therefore,people pay more and more attention to the examination and diagnosis of lung diseases.Elderly people over60 years old have an extremely high probability of chronic obstructive pulmonary disease(COPD)and lung diseases such as pneumonia,even these two diseases occupy the top two in the prevalence and mortality of respiratory diseases.Therefore,the diagnosis of the above two diseases is of great research significance.In recent years,as the field of signal processing continues to advance,speech recognition technology has been widely used.Therefore,this thesis realizes the diagnosis of lung diseases by recognizing breathing signals.Aiming at the problem of noise in different frequency bands in the process of breathing sound signal acquisition,a dual-threshold Empirical Mode Decomposition(EMD)noise reduction algorithm based on a composite evaluation index of root mean square error and smoothness is proposed.First,the breathing sound signal is divided into several Intrinsic Mode Function(IMF)through the EMD boundary;secondly,the boundary threshold is obtained by the root mean square error and smoothness of the components,and then the high and low frequency threshold judgment conditions are used to distinguish High-frequency component,signal component and low-frequency component;finally,reconstruct the signal component to get the noise-reduced breathing signal.To solve the problem of low recognition rate of respiratory acoustic signals,a feature extraction algorithm based on the fusion of Mel-Frequency Cepstral Coefficients(MFCC)and Short-term Energy based on Hilbert-Huang Transform(HHT)was proposed,HHT-MFCC+Energy.First,Hilbert marginal spectrum and marginal spectrum energy were calculated by HHT.Secondly,the spectral energy was used to get the eigenvector through the MEL filter,and then the logarithm and Discrete Cosine Transform(DCT)were used to get the HHT-MFCC feature.Finally,the short-time energy characteristics of breathing sound signals are calculated and fused with HHT-MFCC features to obtain a new feature fusion algorithm.By collecting respiratory signals of related diseases and using an experimental platform to verify the effectiveness of the proposed algorithm,Support Vector Machine(SVM)is used to construct a partial binary tree classification model in the decision binary tree classification method.Experimental results show that the proposed algorithm can effectively identify breathing signals of health,chronic obstructive pulmonary disease,and pneumonia.Precision and Recall are both 0.920,and accuracy is 0.942,which basically realizes the use of breathing sounds.Signals the goal of diagnosing lung diseases.