一個使用智慧型手機的可攜式即時心電圖辨識系統

A Portable Real-time ECG Recognition System Based on Smartphone

顏子豪(Tzu-Hao Yen)

指導教授 ： 余松年

國立中正大學/工學院/電機工程學系/碩士(2013年)

摘要

本研究目的在於建立一套基於智慧型手機的即時心搏辨識輔助診斷系統，以
輔助醫師診斷心臟疾病。傳統心電圖機只能觀測特定時間之心臟電氣活動，瞬間
發生的心律不整不一定能被完整記錄。因此，開發可攜式即時辨識系統渴望成為
上述問題之解決方法。
本系統主要是在Android 帄台上實現無線藍芽心電圖的感測與處理系統，本
系統可分為訊號擷取、R 點偵測、特徵擷取、類神經網路分類及辨識結果呈現。
測詴訊號使用被廣泛認定之MIT-BIH 資料庫，搭配輸出(I/O)將數位資料轉成即
時的類比方式輸出，然後利用低功率的MSP430FG4618對ECG訊號做A/D轉換，
之後透過藍芽晶片與手機做資料傳輸，Android 手機能即時收到12-bit 心電圖訊
號。R 點偵測方面，首先使用帶通濾波器移除雜訊對訊號的干擾，接著採用改良
過的即時R 點偵測演算法找到R 點，取出64 點的QRS 區段後，再使用高階統
計特徵以描述五階小波轉換後之特定次頻帶，共計算27 個高階統計特徵，再搭
配4 個RR interval 頻域相關特徵。此外，我們測詴4 種over-sampling 方法搭配
二次式分類驗證類神經網路效能，選出最佳樣本比例分配之權重，然後將此演算
法移植到智慧型手機上。
結果顯示本系統可以高達97.46%的辨識率分辨七種心電圖，且手機上實際
辨識一個心搏只需0.03825 秒，每10 秒更新一次分析資訊畫面延遲時間低於0.5
秒，顯示本研究演算法的高效能及即時系統的可行性。

關鍵字

即時辨識 ； 高階統計 ； 離散小波轉換 ； 心電圖 ； 智慧型手機

並列摘要

This paper proposed an smartphone-based real-time ECG monitoring and
recognition system toassist the physicians in heart disease diagnosis. ECG
measurement usually requires the patients to carry a device, The recorded ECG
signals are then brought back to the hospital to be examined by the physicians. This
process would take a long period of time and some mistakes or ignorance of minor
signs could be made. These issues give rise to the requisite of portable ECG recording
and recognition system.
This system use bluetooth to receive ECG data and computing algorithms on the
Android platform. It is divided into some functional blocks, include signal acquisition,
R-point detection, feature calculation, classification and result display on the screen
ECG data from the highly recognized MIT-BIH database were selected as the test
signals. The data were transformed in to analog signals using the I/O card. Then the
low power MSP430FG4618 module was used to perform the A-to-D transform. The
digitized 12-bit data were transmitted to the smartphone through Bluetooth. We
remove the noise through a bandpass filter. Then, the improved R-point localization
algorithm was used to locate the R points of the heartbeats. A 64-point QRS segment
centered at the R point was extracted. A five-level discrete wavelet transformation
was used to decompose the segment into different subband components. 27 features
calculated from higher-order statistics were extracted based on the components. Four
RR-interval related feature were also used. In addition, Four over-sampling profiles
combined with the proposed two-stage classifier were tested to verify performance of
the algorithm, The optimal weights were downloaded onto the smartphone as the final
version of the real-time classifier.
This system achieved a high accuracy of 97.46% in identifying seven heartbeat
types on the smartphone, The heartbeat types were recognized in real-time; only 3.825
ms was required to identify a heartbeat. The portability, real-time processing, and high
recognition rate of the system demonstrate the efficiency and effectiveness of the
device as a practical computer-aided diagnosis (CAD) system.

並列關鍵字

real-time recognition ； Android ； Discrete wavelet transformation ； Electrocardiogram ； Higher-order statistics

參考文獻

ges/stat/default.aspx, January 2012

2001-2015 Forecast and Analysis”, Mar 2001.

[3] L. Wei, H. Sheng, S. Zhenzhou, and T. Jindong, “A real-time cardiac arrhythmia

classification system with wearable electrocardiogram”, 2011 IEEE International

Conference on in Cyber Technology in Automation, Control, and Intelligent