以穿戴式心電圖收集刺激活動下之心率資料

Abstract

在過去，要量測心電圖需親自到醫院或診所透過專業的設備進行量測，時間跟地點皆受到了限制。然而近年來拜微機電系統演進之賜，許多廠商開發了可攜性非常高的穿戴式心電圖，讓我們得以在各種情境下進行量測。但大部分的穿戴式心電圖並無經過美國食品與藥品管理局(FDA)的認證，因此無法保證其訊號品質。而我們希望利用傳統偵測醫療等級ECG設備的QRS偵測演算法來驗證其是否可成功辨別穿戴式心電圖資料，我們實作了三種演算法，包括小波轉換QRS偵測演算法、Pan & Tompkins演算法及基於移動平均濾波器之QRS偵測演算法，並透過穿戴式心電圖收集了跑步機以及慢跑時的心電圖資料來驗證演算法，偵測的結果精確度皆大於92%，召回率則在90%左右。 此外，我們亦利用穿戴式設備的可攜性來研究情緒變化（刺激、緊張）與心率加速之間的關係。我們選擇了兩個會造成情緒起伏較大的情境下進行實驗，分別為看電影以及遊樂園。目的是藉由分析心率資料判斷受測者的緊張刺激程度，因此我們設計了一個用來標記出刺激片段的演算法(MES演算法)，並針對兩個情境設計了問卷，調查該項目真實的刺激程度。實驗的結果證明了心率的大小與刺激、緊張的程度是呈現正相關，且MES演算法所標記之刺激片段與問卷之趨勢非常接近。

The Electrocardiograph (ECG) must be measured by professional equipment in hospital or clinic in the past, which has restrict of time and place. However, in recent years, thanks to the innovation of MEMS, many companies have developed the wearable ECG which is highly portable. But until now, most of the wearable devices for physiological monitoring haven’t acquire the FDA approval, and the measured data may have some level of noise. Hence, we hope to prove if traditional QRS detection algorithm is feasible for wearable ECG data. We implement three classic QRS algorithms and develop a mobile app to collect wearable ECG data in stimulating situations (treadmill and jogging). And to verify whether the algorithms can precisely discriminate QRS complexes in these two situations. The results show that all of the precisions are more than 92%, and the recalls are about 90%. On the other hand, with the portability of wearable ECG, we select two situations (movie and amusement park) which may cause people excited and nervous, and conduct the heart rate measurement0 to research the correlation between emotion and heart rate. Our purpose is to assess the exciting level by analyzing heart rate data. We design an algorithm (MES algorithm) to mark the exciting segments, and design questionnaires to investigate the real exciting level. The experiment results prove that the value of heart rate and the exciting level have a positive correlation, and the results of exciting segments marked by MES are close to the results of questionnaires.