應用時頻分析於建立運動員肺音訊號比較系統

Abstract

胸部聽診一直以來都是醫生用來診斷病人肺部病症的主要方法，醫生藉由聽診器聽取肺部的聲音，來判別病人的健康狀況以及所得病症。但自從科技的進步，數位聽診漸漸取代了傳統聽診，成為新一代主流的技術。數位聽診具有多工處理，可記錄，以及可數據化演算之優點，使得聽診技術進入一個新的紀元。透過將肺音訊號影像化，以及使用科學上訊號分析的訊號法來分析肺音訊號，我們可以從肺音訊號得到比已往更多的訊息。本研究的目的為透過肺音擷取系統以及時頻分析方法來建立一套可以觀察不同運動種類運動員之肺音訊號之系統，並透過此系統來分析不同運動種類以及其肺音訊號之間的特色與關係。 首先使用一般聽診器與數位單指向性麥克風（鐵三角AT9904）做結合，製作成16個數位聽診器，並在人體背後肺部的位置布置麥克風陣列。透過USB音效卡以及Audio Stream Input Output(ASIO)驅動程式將肺音訊號錄製至電腦硬碟中並儲存。儲存在電腦中的肺音訊號，透過MATLAB以及Visual Signal軟體做訊號前處理以及時頻分析。本研究採取了三種時頻分析方法，分別為短時傅立葉轉換法（Short Time Fourier Transform，STFT）、Morlet小波轉換法（Morlet Wavelet Transform，Morlet WT）以及希爾伯．黃轉換法（Hilbert Huang Transform，HHT）。三種分析法各有其特色以及優點，但三種分析法中，本研究認為HHT最適合做為不同運動種類運動員之肺音差異之比較。熵(Entropy)之原理也用來作為呼吸氣流流速模型的參數之一。本研究透過以上方法來建立一個可以觀察不同運動員肺音特色之系統

In the diagnosis of the respiratory diseases, auscultation is a main method to diagnose and treat patients. Auscultation is a non-invasive and convenient diagnostic method. With the upgrading of computer hardware, Digital auscultation is gradually replacing traditional auscultation. Auscultation digitizing overcomes the shortcomings that are not sensitive and the subjectivity of the human’s ear. And auscultation also has the advantages that are multi-tasking, recordable and can be analyzed by computer. We can know a lot of information by image-based of lung sound signals. In the digital auscultation analysis, what method I use to analyze the lung signals which microphone recorded will affect the results of the experiment greatly. The purpose of this study is to build a system to observe the lung sounds of the different athletes by combination of lung sound recording system and time-frequency analysis. And find the lung sound characteristics and relationships of different athletes from the results. First, I combine the stethoscopes and unidirectional microphones (Audio-Technica AT9904) as 16 digital stethoscopes, and arrange the digital stethoscope array on the back of human’s body. Through the USB sound cards and Audio Stream Input Output(ASIO) driver, the lung sound signals are recorded and saved in the hard disk. The signal would be processed and analyzed by MATLAB and Visual signal. The research has used three time-frequency analysis methods, they are Short Time Fourier Transform (STFT), Morlet Wavelet transform (Morlet WT) and Hilbert Huang transform (HHT). This research considers the result of HHT is the most suitable for lung sound signals analysis. Entropy is also used for respiratory air flow detection. This research build a system to observe the characterizes of lung sound signals from different athletes by the above method