摘要
每個人的身體特徵都是獨一無二的，且不易被人盜用，因此近年來，生物辨識 (biometric identification)已經大量應用在生活中，像是人體指紋、臉部、聲音或虹膜等，來辨識使用者身分。光電容積圖 (PPG)隨著生物傳感器 (biosensor)技術的提升，量測到的訊號也越來越精準，再加上只要利用穿戴裝置就能有效率地進行量測，因此利用光電容積圖去實行生物辨識也是未來的趨勢。 在本篇論文中，我們提出一套能在行動裝置上運行的生物辨識系統架構 (biometric identification system)。除此之外，如何從光電容積圖抓取具有獨特性 (unique)的特徵 (feature)，也是我們在本篇提出來的重點。在我們提出的生物辨識系統架構包含識別 (identification)和驗證 (verification)兩個分類階段，並以深度學習為核心去實作兩個階段的分類器 (classifier)。
並列摘要
Photoplethysmographic (PPG) signal, also known as PPG signal, can be used as a biometric technique due to its unique form that differs between individuals. Rapid advances in miniatures of PPG sensors and smart bracelets have enabled the proliferation of low-cost, wearable biosensors capable, which makes PPG have great potential to serve as a biometric identification system. In this thesis, a wearable human identification system based on PPG signal collected from a wearable sensor on a smart bracelet is introduced. To improve the performance of the human identification system, we utilize a deep neuron network which takes raw data and preprocessed features of collected PPG signal as input. In addition, we propose a human identification architecture to offer biometric identification and biometric verification. The proposed biometric identification scheme has been tested on PPG databases and experimental results show that our method achieves higher recognition accuracy than a set of existing methods.