對於一般肺部疾病，醫師最先採取的診療方式，就是使用聽診器是來聽取肺音訊號。但由於器材與環境的限制，肺音常受到雜訊的干擾。其中，心音是主要的干擾來源。由於肺音與心音並不是線性非時變(Linear and Time-Invariant, LTI)且不是平穩(Stationary)訊號，所以若使用傳統的傅立葉轉換(Fourier Transform)來進行分析，將會無法獲得正確的資訊。本論文使用黃鍔博士在1988年所提出的希爾伯特-黃轉換(Hilbert-Huang Transform)理論，這種轉換特別適合用來處理非平穩的訊號。透過經驗模態分解(EMD)，將訊號分解為一組內蘊模式函數(IMF)；內蘊模式函數通常由高頻至低頻被分離出來。接著，對每一個內蘊模式函數分量進行希爾伯特轉換(Hilbert Transform)，得到頻率對時間的瞬時變化。希爾伯特頻譜(Hilbert Spectrum)在時域以及頻域具有良好的分辨率，故在此使用希爾伯特-黃轉換，來分析心音與肺音訊號，藉以降低心音對肺音的干擾；幫助醫療人員在診斷上，有更好的判斷與觀察。In this research, we take heart sound signals as interference to lung sounds and propose a method to reduce the interfering heart sounds in lung sounds. The lung sounds were obtained by placing an electronic stethoscope head on the chest of the subject and recording the output signal of the microphone in the stethoscope head. We incorporated Hilbert-Huang Transform (HHT) in our heart sound reduction. HHT was proposed by Norden E. Huang. It is especially suitable for processing non-stationary and non-linear signals, such as physiological signals. In HHT, the target signal can be decomposed into a number of intrinsic mode functions (IMFs) by empirical mode decomposition (EMD).These IMFs can be transformed into the Hilbert space, and then their instantaneous frequencies can be observed in the time domain. The performance of our heart sound reduction algorithm was evaluated in terms of the heart-sound-noise reduction percentage (HNRP), which .is about 80% in our experiments. This result is comparatively better than that of a wavelet-based method shown in the literature.
