用於行動照護應用之低能量同步非同步混合式心電訊號特徵擷取器設計

An Energy-Efficient Mixed Sync-Async Cardiac Delineator for Mobile Healthcare Applications

張博堯(Chang, Poyao)

指導教授 ： 李鎮宜

國立交通大學/電機學院/電子工程系所/碩士(2012年)

https://doi.org/10.6842/NCTU.2012.00121

摘要

行動照護應用使用了具備無線傳輸能力的感測器來達到生理狀況監控，能夠長時間的監控就成為了這類應用的主要需求。我們藉由在感測器上做訊號處理以減少無線傳輸的資料量。在感測器上做事先的特徵擷取，傳輸的資料將可以被大大地減少，同時擷取出的生理特徵也可被用於即時的疾病診斷，減少感測器到醫院的診斷延遲時間，給予病患更大的保護。
以心血管疾病為例，心臟疾病的診斷是藉由心電圖的P、Q、R、S、T等特徵的大小及區間來判斷。在這份論文中，我們提出了一個基於小波轉換的心電訊號特徵擷取演算法及硬體，並且使用兩個標準的心電圖資料庫來驗證特徵擷取的結果。我們提出的演算法對所有提供的特徵都達到了99.4%及96.1%以上的靈敏度及準確率。在硬體架構上，擷取器被計在極低的操作頻率（250Hz），並藉由共用搜尋核心、記憶體最佳化、以及觸發式的獨立電源管理非同步搜尋核心，來達到低能量消耗的需求。極低的操作頻率加上非同步電路更提供了降低工作電壓來降低功率消耗的可能性。我們提出的心電訊號擷取器採用90奈米標準CMOS製成晶片實現。在0.5V 的電源供應之下，擷取器功率消耗為2.56μW。最後，我們也利用了市售晶片以及嵌入式處理器建構了一個極小的無線傳輸感測器模組來驗證我們的特徵擷取演算法。藉由這個感測器模組，我們驗證了演算法的行動環境中的準確率以及藉由特徵擷取減少資料傳輸能量的想法。

關鍵字

心電圖 ； 特徵擷取 ； 低功耗 ； 非同步電路

並列摘要

Long-term monitoring is the key requirements for mobile healthcare applications, where the wireless sensor nodes are worn to record the human’s vital signals. On-sensor signal analysis is proposed for these applications to enable timely detection of risky syndromes and extend the monitoring time. Instead of raw data transmission, the wireless transmission energy is reduced by only transmitting the vital features.
In case of cardiac diseases, the syndrome analysis is performed based on different extracted features of ECG signals like P, QRSon, R, QRSend, and T wave. In this work, an energy-efficient cardiac delineation algorithm based on multi-scale wavelet transform is designed together with its hardware implementation. The detection result is evaluated on two annotated databases including MIT-BIH arrhythmia database and QT database. The obtained sensitivity and positive predictivity are over 99.4% and 96.1% for the five ECG features, respectively. With shared search kernels, storage optimization and event-driven asynchronous search kernel with individual power management, the delineator can operate at 250Hz without the needs for additional high speed clock. The slow operating speed and asynchronous search kernel also enables further voltage scaling to reduce power. Implemented using UMC 90nm technology and operating at 250Hz with 0.5V supply voltage, the overall power is 2.56μW for real-time ECG monitoring.
Besides, a miniaturized prototype wireless sensor is constructed using commercial products with on-sensor delineation. The prototype provides evidence to the delineation robustness for mobile monitoring and power reduction for on-sensor feature extraction.

並列關鍵字

ECG ； Feature Extraction ； low power ； asynchronous circuit

參考文獻

[2] Malcolm S. Thaler, “The Only EKG Book You'll Ever Need 5th, ed.”, Lippincott Williams & Wilkins, 2009.

[3] Goldberger AL, Amaral LAN, Glass L, Hausdorff JM, Ivanov PCh, Mark RG, Mietus JE,et.al., PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation 101(23):e215-e220 [Circulation Electronic Pages; http://circ.ahajournals.org/cgi/content/full/101/23/e215]; 2000 (June 13).

[4] Moody GB, Mark RG. “The impact of the MIT-BIH Arrhythmia Database,” IEEE Eng in Med and Biol 20(3):45-50 (May-June 2001).

[5] Laguna P, Mark RG, Goldberger AL, Moody GB., “A Database for Evaluation of Algorithms for Measurement of QT and Other Waveform Intervals in the ECG,” Computers in Cardiology 24:673-676 (1997).

[6] V. X. Afonso, W. J. Tompkins, T. Q. Nguyen, and L. Shen, “ECG beat detection using filter banks,” IEEE Trans. Biomed. Eng., vol. 46, pp. 192–202, 1999.