Research On Respiratory Sound Data Acquisition And Quantitative Analysis

With the aggravation of the air pollution and increasingly prominent of the environment issue,respiratory system disease becomes a hot topic of public concern.Along with it,the public begin to pay more attention to the respiratory system disease.In the meantime,research of the respiratory sound signal becomes a hot topic and focus that the society pays close attention to for a large number of physiological and pathological information of the respiratory system it contains.As a non-invasive monitoring method,respiratory auscultation has been used widely.However,because of the clinical common of the respiratory system diseases are mixed lesions and the locations of lesions are not fixed,traditional auscultation in single channel cannot provide a wealth of clinical information.Therefore,based on the results of the previous research of our laboratory,in this paper,we propose a multiple locations collecting method for the respiratory sound acquisition and research a set of fast and effective algorithms after a deep analysis of the signals collected form multiple locations to distinguish the respiratory sound signal is normal or abnormal for the respiratory sound collected from the triangle location under the cricoid cartilage.In order to complete the above objectives,this thesis mainly studied the following aspects:(1)New design method and acquisition for the respiratory sound collectingAfter an in-depth research and study on the relationship between the respiratory sound and respiratory diseases,we put forward a new collecting method-collecting in 7 parts(triangle location under the cricoid cartilage,left upper lung,left middle lung,left lower lung,right upper lung,right middle lung and right lower lung)-for the respiratory sound.Until now,the collected data includes: 54 normal respiratory sounds from 25 volunteers without respiratory disease in Xihua University,46 abnormal respiratory sounds,in which includes 30 asthma respiratory sounds from 20 patients with respiratory disease in the respiratory medical ward of Chinese People's Liberation Army 452 Hospital.The storage of the respiratory sounds used for analysis was completed successfully.(2)Pre-processing of respiratory sound signalsIn this section,we mainly focus on the noise source analysis of the collected respiratory sounds and design the corresponding processing approaches for different respiratory sounds such as Band-pass Filter denoising,Adaptive Wavelet Threshold denoising(AWT)and Stationary – Non-Stationary Filtering technology(ST-NST).Through the above methods,we would like to improve the signal-to-noise ratio of respiratory sounds.(3)Feature extractionAccording to the characteristic of the respiratory sound,we combine the time-domain analysis with the frequency-domain analysis to extract the characteristic features.In time domain,we mainly use the Homomorphic filter to extract the envelope of the respiratory sound and extract the switchover points of expiratory phase and inspiratory phase,the duration time of expiratory phase T1 and inspiratory phase T2,the ration of T2 and T1(Tr=T2/T1)and the respiratory frequency Rr.While in frequency domain,we employed the Fast Fourier Transform(FFT),frequency band characteristic extraction based on wavelet transform and the frequency of area quartering in frequency spectrum to extract the normalized energy value of each frequency bands(T=[E1 E2 E3 E4 E5 E6 E7])and the frequency of area quartering in frequency spectrum(Q25 ? Q50 ? Q75).Based on these parameters,we conduct the preliminary identification for respiratory sound of normal person and asthmatic patients.(4)Quantitative analysis of clinical respiratory soundBased on the feature extraction method mentioned above,we extracted the characteristic parameters of the respiratory sound collected in clinical,made a data survey and analysis according to the research of characteristic parameters and built a preliminary quantitative analysis system of clinical respiratory sounds.Furthermore,we have collected respiratory sounds from two asthma patients during all of their treatments and made an analysis for their respiratory sounds collected at the moment that they were treated in the hospital and leave the hospital.This thesis mainly researched the respiratory sounds of normal people and asthma patients,built a preliminary quantitative analysis system of clinical respiratory sounds.Then verify this system.Furthermore we have collected respiratory sounds from two asthma patients during all of their treatments.We can know the treatment effects according to the visual datas.