本文運用EdgeTPU為核心，透過PoseNet偵測人體的姿勢，而PoseNet在偵測關節點時會有不穩定之現象，為了解決此問題，本文運用到卡爾曼濾波器，將PoseNet測量結果進行濾波，進而達到訊號輸出更加穩定，更能準確的判斷及觀察患者的當下姿勢，而人工智慧的應用已遍布社會各個行業，如本文運用PoseNet技術在病床上方，偵測該病患是否有離床現象，以降低人力資源，給予醫師與護士醫療上的便利性與降低人力資源的消耗。

This article uses EdgeTPU as the core to detect human posture through PoseNet. PoseNet is a bit unstable when detecting joint points. In order to solve this problem, this paper uses Kalman filter to filter the PoseNet measurement results to make the signal output more stable and to more accurately judge and observe the patient’s current posture. The application of artificial intelligence has spread to all walks of life in society. For example, this paper uses PoseNet technology to detect whether the patient has gotten out of bed in bed to reduce human resources, provide medical convenience for doctors and nurses, and reduce the consumption of human resources.