基於FPGA的快速哮鳴音偵測系統

Abstract

哮鳴音是許多呼吸道疾病診斷的指標，哮鳴音的偵測可以協助醫生對這類患有慢性呼吸道疾病的病患進行長時間的監測，以預防緊急狀況的發生。本文提出一個使用現場可編程邏輯閘陣列(Field Programmable Gate Array, FPGA)的快速哮鳴音偵測可攜式平台，利用FPGA可以硬體加速來達到快速偵測的特性，更可以讓我們的哮鳴偵測系統能夠進一步地製作成系統晶片(System on Chip, SoC)，甚至能與其他生理訊號偵測系統整合成更先進及更複雜的系統晶片。 本系統首先將聲音切割成每兩秒為一個處理單位，並藉由短時傅立葉轉換(Short-Time Fourier Transform, STFT)來求出哮鳴音的時間與頻率上的相關性，並針對時頻圖做雙邊濾波(2D Bilateral Filter)、邊緣偵測(Edge Detection)、多閾值影像分割(Multi-threshold Segementation)、影像形態學處理(Morphological Processing)、以及影像標記(Image Labelling)來萃取出哮鳴音特徵，並以支持向量機(Support Vector Machines)對哮鳴音的特徵進行分類訓練，我們便可以利用訓練好的支持向量機模型對哮鳴音及正常呼吸音進行辨識。 在使用Xilinx ML605開發平台實現後，本系統可以達到51.97MHz的處理速度，偵測系統的效能(Performance)達到0.912，使我們可以在短時間內偵測哮鳴並達到快速監測哮鳴的目的。

Wheezes have often been treated as an important indicator to diagnose the obstructive pulmonary diseases. A rapid wheezing detection system may help physicians to analyze and to long-term monitor the patients’ situations. This thesis proposes a portable wheezing detection system based on Field Programmable Gate Array (FPGA). It accelerates wheezes detection. It could flexibility function as a single process system or be integrated with other biomedical signal detection system. Firstly, the sound signal is segmented into units of 2 seconds. Then short-time Fourier transform was used to look into the relationship between the time and frequency components of the sound data. Thereafter, we continued processing the spectrogram by 2D bilateral filtering, edge detection, multi-threshold image segmentation, morphological image processing and image labeling to extract the wheezes features according to Computerized Respiratory Sound Analysis (CORSA) standards. Then these features were used to train Support Vector Machines (SVMs) and built the classification models. Finally, this trained model is used to distinguish to detect wheeze for new coming sound data. This system runs on Xilinx ML605 platform. Experiment results show a high performance of 0.912 in analysis of wheeze recognition in hardware. The detection process is good for 51.97 MHz clock frequency. It is good for high speed classification for wheeze.