In order to solve the undiscovered problem that caused by the fall of  elders, we develop a deep learning based system and named it 「SkyEye」,which includes our own sensor 「AI Falling Image Sensor」, a cloud sever that stores and pushes information about fall events and a mobile app that communicates with users.

1.跌倒檢測演算法：
在預測跌倒的部分，本系統分別使用了脊椎線(Torso line)的偏移角度、加速度與人體長寬比、對應物體速度之影像距離測量、長短期記憶(Long Short-Term Memory, LSTM)的RNN等方法來建構跌倒檢測演算法。經過實際的參數調整與測試，其辨識跌倒準確率可達92%，辨識每秒張數(FPS)在8~9之間，其速度可達到即時處理(Real-time)之效果。
2.系統整合「SkyEye」：
本團隊自製研發的AI跌倒影像感測器，其感測結果能夠實現系統化，並結合雲端伺服器將老人的跌倒地點、時間、狀態等，推播至家屬或看護人員的手機APP，能做到即時通報的功能。此外，雲端伺服器也能夠儲存AI跌倒影像感測器所偵測到的人體關節點，有利於老人跌倒的了解及評估，藉此提升跌倒檢測效果。

With system integration provided by 「SkyEye」, Mobile phone app for family members or caregivers can instantly be reported by the cloud server when the fall event happened, that can improve the efficiency of the nursing station to grasp the information of the elderly fall, and reduce both rescue time and chance of death of the elderly.