| Information mechanism of respiratory diseases in the respiratory system reactions often scattered distribution throughout the chest area, breath sounds auscultation breath sounds as a major means of monitoring, has been more widely used, but because of the breathing sound clinical common disease mostly mixed lesions traditional single lead only by repeated auscultation auscultation different parts in order to provide clinical information, and multi-channel acquisition breath sounds at the same time conducive to different parts of the breath sounds and murmurs between the duration, intensity and other information collected, providing a wealth clinical information, and the acquisition of the respiratory signal for further analysis. To accomplish these objectives, the paper-depth research in the following areas:(1) be familiar with basic information breathing sound, grasp the mechanism and the relationship between patients breath sounds breath noise and respiratory diseases; and methods of study and research abroad breath sounds, and found problems.(2) analysis and laboratory cooperation of Yamaguchi University has independently developed the first Meets breath sounds auscultation collection, this chest-Technica AT9904 microphone and mainly by the 3M company produces shock chamber(Littmann, Classic IISE) composition.(3) The combination of Chengdu 452 Hospital physician guidance, five lead collection site, the location and treatment method based on auscultation has applied for a utility model patent is now using this program collected more than 50 cases of respiratory sounds 20 normal volunteers breath sounds and 21 cases of abnormal respiratory sounds 20 patients(14 of which have the same pathology), completed a preliminary breath test tone collection.(4) of the collected signal preprocessing breath sounds analysis, design the appropriate treatment for a variety of noise, such as: frequency notch filter, wavelet threshold noise, IIR filter design, as much as possible by the above program improve the SNR of respiratory sounds, by contrast experiment, the pretreatment effect is more obvious.(5) The envelope extraction and feature extraction, for the characteristics of breath sounds, using the first analysis, supplemented by the frequency domain, and then mainly to extract the envelope in the time domain, contrasting common envelope extraction method based on Hilbert Transform breath sounds envelope extraction, based on a normalized average Shannon energy envelope extraction breath sounds, breath sounds packet-based network to extract a single degree of freedom model, based on Morlet wavelet breath sounds envelope extraction of comparative tests, select Morlet package envelope extraction algorithm, then by using FCM algorithm envelope wave threshold line division, and finally get the characteristic parameters expiratory and inspiratory phase duration(T1, T2), expiratory time and inspiratory pause time gap(D1, D2), and peak expiratory inspiratory phase(P1, P2), with T1 / T2, D1 / D2 as the distinction between types of parameters breath sounds, a preliminary breath sounds dichotomous.(6) The ultimate goal of this laboratory is to be able to lower acquisition, support the diagnosis of back-end server, and then how to use fewer resources to storage and transmission progressively increased database and upgrade a single acquisition data transfer speed is required article a research direction, experiments to override OMP algorithm and SAMP algorithm to breath sounds, through the steps and conditions of the two algorithms required during operation involved comparing the results required under the premise of unknown sparsity(K), and improved algorithms and calculation accuracy speed up computations. In summary, this paper for the collection and analysis of respiratory sounds were studied, and the use of actual clinical data collected to validate results show: the acquisition of five guides breath sounds auscultation advantages with respect to single-lead acquisition significantly, it could also be collected After more intuitive data for different parts of the data comparison and recognition, but also on the breath sounds were better feature extraction classification based FCM clustering algorithm, and the future focus will be on increasing the variety of abnormal breathing sound judgment and optimization algorithm breath sounds. |
