以光體積描述訊號及三軸加速器之抗晃動心率監控

Abstract

在臨床醫療中，對醫生而言心跳速率是一項重要的生命跡象來診斷人的基本健康狀況。心跳速率會根據各種不同的生理現象而產生變化，如心律不整，焦慮等等。更重要的是，正常的心跳速率因人而異。因此充分掌握個人心跳速率狀況是一項很重要的健康評估指標。除了在臨床醫療中，在運動當中即時監控心跳速率能夠有效的幫助我們達到最大的運動效率以及有效地防止受傷。根據上述的好處，近年來心跳速率監控器的需求越來越多的趨勢。 即時的心跳速率監控應用中，傳統的心跳速率監控是以心電圖來量測，然而心電圖的成本高及其使用上有所不便。目前心跳速率監控的新趨勢是利用光體積描述訊號，其有成本低，使用方便及能整合在穿戴式裝置上。然而光體積描述訊號在使用上遭遇最大的困難為晃動(Motion Artifacts)的干擾，過去相關的解決方法應用在實際上並不可行導致結果不準確。因此，我們需要有額外的訊息來分析訊號當中屬於晃動所造成的成分。三軸加速器的訊號能夠幫助分析晃動成分並且在受干擾的訊號當中抑制該晃動成分。在這篇論文中，我們提出了以光體積描述訊號及三軸加速器建立的抗晃動心跳速率監控系統。該系統具有兩個重要的部分，晃動消除和頻譜峰值追蹤。我們提出了一個基於可適性雜訊消除的低複雜度晃動消除法來有效地抑制晃動成分和基於卡爾曼濾波器的穩固頻譜峰值追蹤法來追蹤心跳在頻譜上的峰值。這個系統在心跳速率監控上有很好的準確度，展現出實際應用在穿戴式裝置上的潛力。

In clinical medical, heart rate (HR) is an important vital sign for doctors to access the general health of a person. Depending on the various physiological condition, like arrhythmia, anxiety and so forth, the HR changes accordingly. What’s more, normal HR varies from person to person. Knowing yours can be an important heart-healthy gauge. In addition to clinical medical, monitoring HR when exercising can help us to maximize the effect of exercising and prevent from getting injured. Benefits from the advantages mentioned above, the demands for HR monitors have increased in the recent years. In the application of real-time HR monitoring, Electrocardiogram (EKG) is traditional ways to monitor heart rate. EKG encounters high cost and inconvenience. Recent trend for HR monitoring is Photoplethysmography (PPG), which is low-cost, user-friendly and can be integrated into wearable devices. However, PPG will be interfered by motion artifacts (MA). Past works for HR monitoring from motion-corrupted PPG is not workable in real applications. As a result, we require additional information to analyze the signal components corresponding to MA. Tri-axis accelerometer is helpful to analyze the components of MA and suppress them from the corrupted PPG signal. In this work, we propose a motion-tolerant HR monitoring system with PPG signal and tri-axis accelerometer. The system has two key parts, motion artifact cancellation and spectral peak tracking. We propose a low-complexity motion artifact cancellation based on adaptive noise cancellation (ANC) to suppress the MA components effectively and a robust spectral peak tracking based on Kalma filter to track the peak corresponding to HR on the spectrum. This work has good accuracy of HR monitoring and show the potential to apply on wearable device in real application.