視訊生理信號擷取之探討

Abstract

心律在大部分的情況下，是一個判斷身體狀況的重要指標，藉由測量這個參數，我們可以瞭解目標的活動情況，在過去的研究中，人們開發了基於視頻的心跳速率測量方法，其原理大致為利用心臟在運動的時候，皮膚表層的血管會擴張和收縮，造成皮膚顏色發生一些肉眼觀察不到的變化，同理，瞳孔周圍的微血管也會有相同的變化，透過放大這些訊號，偵測使用者的心率，為了測試此種測量方法對於更高的心率以及追蹤心率變化的可能性，本研究將此種藉由視頻的測量方法，進一步應用到老鼠以及運動中的人上。 在研究方法上，我們分別對人以及麻醉中的老鼠，測量其穩定狀態下的心律，在追蹤心率變化的實驗上，我們讓人類受測者在室內運動之後休息，使其心率產生上升以及下降的變化，在老鼠的實驗中，我們利用影響心臟速率的兩種藥物（異丙腎上腺素和三卡因）進行分別的測試。 在估計心率的方法上，我們分別對不同顏色的訊號，經過移動平均以及帶通濾波器等方法濾除雜訊，將處理過後的訊號藉由波峰偵測估計可能的心律，最後將實驗結果與實際心電圖比較並計算誤差，並試圖比對信號找出誤差的來源。 實驗結果在靜態的實驗上表現較好，人的誤差平均大約在2%以下，這代表心跳在平均1分鐘60的狀況下，我們的方法大約只有不到1下的誤差，老鼠的誤差平均大約在3.5%左右，經過調整移動平均和波峰偵測的方法，誤差可以下降至2.37%，在心律變化的實驗中，我們可以對一個短時間(5秒~10秒)的心率有一個較準確的估計，然而要準確估計每一下心跳的準確時間是有誤差的，最後，本研究提供一種相較於ECG更為簡單且方便的方法來追蹤人或老鼠的心率變化，並在未來可以進一步開發為穿戴式裝置或應用在老鼠實驗上。

In most cases, heart rate is an important indicator of the physiology condition. By measuring this parameter, we can understand the activity of the target. In the past study, people developed a video-based heart rate measurement method. The theory of this method is that when heart is active, blood vessels will expand and contract, resulting in skin color varies slightly with blood circulation. This variation, while invisible to the naked eye, can be exploited to extract pulse rate. This heart rate measurement method without using needle electrode will allow the patients to feel comfortable in the process. In order to test the possibility of this measurement method for higher heart rate frequency and tracking changes in heart rate, this study will be applied to the rats and human target. In the study, we measured the heart rate in the steady state of the human as well as the rats in the anesthesia. In the experiment of tracking the change of heart rate, we let the human subjects rest after the indoor exercising, so that the heart rate would rise and then fall during the procedure. In the mouse experiment, we used the two drugs (methylatropine vagal blocker and propranolol beta blocker) that affected the heart rate were tested separately. In the method of estimating the heart rate, different color signals were applied moving average and bandpass to filter out the noise. Then we estimate the heart rate of the processed signal. At last, we compared the actual electrocardiogram and estimation result and calculated the error and attempted to find the source. The experimental results performed better in steady state experiments. The average error of human steady heart rate detection is less than 2%. It means that when heart rate is 60bpm, only one or two peaks might be over or under estimated. The average error of rat steady heart rate detection is about 3.5%. After adjusting the moving average and peak detection method, we were able to decrease the error to 2.37%. In the heart rate change experiment, we have a more accurate result of short time (5 seconds to 10 seconds) heart rate. However, it is unlikely to estimate the each heartbeat ta exact time. Finally, this study provides a simpler and more convenient way to track heart rate changes in human or rats than ECG. In addition, this method will able to be applied in wearable device or rat experiment.