學習式非接觸心跳及呼吸偵測

Abstract

非接觸式的心跳及呼吸偵測方式愈來愈受到重視，遠距光體積描記法(rPPG) 僅需靠一般消費性相機錄得的影片中的皮膚顏色變化便能偵測心跳，近期的研究著重在提高心跳造成的顏色變化訊號的強度，方法包括使用獨立成分分析(ICA) 以及研究皮膚色度模型(chrominance-based model)，本論文中，我們將從另一個角度分析這個問題，我們提出學習式方法兼納多種特徵資訊，並且得到長足的進步，利用支持向量回歸(support vector regression) 學習色度模型方法的特徵資訊能使均方根誤差由22.7 降到7.31，同時相關係數由0.30 成長至0.77，若加上我們提出的多特徵結合學習與多片段結合學習方法，可將均方根誤差降至5.48 而相關係數為0.88，這個學習式方法可擴展為包含其他新的特徵資訊。 對遠距人體呼吸起伏追蹤來說，深度攝影機如Kinect 是一個較便宜的選擇，大部分的非接觸式呼吸頻率偵測研究關注靜止的人體，本論文中，我們利用不同身體部位的相對位移來計算呼吸造成的活動中人體胸部起伏，學習式方法同樣被運用在這個問題上，據我們所知，這是第一個嘗試遠距離偵測活動中人體的呼吸頻率的研究。

There is growing interest in non-contact heart rate and respiratory rate detection. Remote photoplethysmography (rPPG) enables measuring heart rate from recorded skin color variations with consumer cameras. Recent research has aimed to improve the signal strength of color variations caused by heart beat by using independent component analysis (ICA) technique or analyzing chrominance-based model. In this thesis, we argue for treating this emerging problem in a novel aspect – proposing a learning-based framework to accommodate multiple and temporal feature and yielding significant and robust improvement. Using support vector regression (SVR) on published chrominance-based feature improves the root mean square error (RMSE) from 22.7 to 7.31 as well as correlation coefficient (CC) from 0.30 to 0.77. With proposed novel multiple feature fusion and multiple segment fusion techniques, we achieved the best estimation result with RMSE 5.48 and CC 0.88. Meanwhile, the proposed framework can be extended to other promising features. Depth camera, like Kinect, is a low-cost choice to remotely track respiratory motion of human body. Most non-contact respiratory rate detection researches focus on stationary person. In this thesis, we use relative displacement of different body parts to find respiratory motion of person in action. Learning-based method used for heart rate detection is also applied here. To our best knowledge, we are the first to remotely detect respiratory rate of person in action.