以快慢可適性梳型濾波器消除心率監測器之運動偽影

Abstract

這篇論文提出一個新穎的消除運動偽影演算法，可應用在心率監測器。我們的演算法不需要額外的硬體，例如加速儀或其它的參考信號。本文使用提出的快慢可適性梳型濾波器(SFACF)去消除光體積變化描記圖(PPG)訊號所耦合的運動偽影。週期的心跳具有諧波成份與心跳連續變化的兩個特性是本演算法的重要依據。由實驗結果可知，此方法對基頻訊號與諧波成份產生了放大的效果。心率監測器每一秒更新一次心跳頻率。我們的方法和可適性雜訊消除濾波器(ANC)與頻譜峰值尋找梳型濾波器(SPS-CF)進行了比較。當量測得到的PPG訊號品質較差的情況下，我們的快慢可適性梳型濾波器，優於其他兩種濾波器。此外本方法使用2015 IEEE Signal Processing Cup的資料庫數據，由實驗結果顯示提出的方法有好的性能，參考心跳頻率與估測心跳頻率兩者之間的R平方值為0.96。最後本方法也和消費性電子產品心跳帶比較，我們估測出的心跳頻率和商業用品有很高的相關性。但是心跳帶配戴時要掛於前胸 ; 而我們的只在手腕上。此外，我們在成本上也有優勢。多位年紀在32~58歲的男、女受測者參與我們的實驗，都得到滿意的結果。

In this dissertation, an algorithm for eliminating motion artifacts in heart rate monitoring by using a photoplethysmography (PPG) sensor is proposed. No additional hardware, such as accelerometers or other reference signals, is needed to apply our algorithm. The proposed slow–fast adaptive comb filter (SFACF) distinguishes motion artifacts in PPG signals based on the harmonic and continuous features of the heart rate. The effectiveness of fundamental and harmonic frequency enhancement in real time is experimentally demonstrated with the heart rate being updated every second. The performance of the SFACF, which does not require an additional reference signal, is compared with the performances of adaptive noise cancellation (ANC) and spectral peak search-comb filtering (SPS-CF). It was determined that, for a low-quality PPG signal, the SFACF performed better than ANC and SPS-CF. Experimental results on the PPG datasets used in 2015 IEEE Signal Processing Cup showed that the proposed approach had excellent performance, and the R-squared between the estimates and the ground-truth of heart rate was 0.96. Finally, measurements conducted using a commercial heart rate monitor and those using our algorithm were shown to be highly correlated. However, a commercial heart rate monitor requires the wearing of a chest strap. On the other hand, our algorithm enables the wearing of a wrist device for heart rate monitoring. In addition, the proposed method can reduce hardware costs. Male and female subjects between the ages of 32 and 58 participated in an experiment we conducted using the proposed algorithm. Based on the results, the proposed SFACF offers a satisfactory performance approach.