運用Kinect於養護中心寢居狀態偵測警示技術的實作

隨著國內老年人口的增加，長者安養與長期照護的需求越來越高。不過時常會看到養護中心的護理人員在夜間人力編制較少的情況下，無法即時發現長者突然下床的狀況，導致長者可能因此而失足跌倒，衍生後續照顧問題或者糾紛。有鑑於此，在本論文中透過Kinect感測器的紅外線攝影機取得深度資料，再利用深度資料進行運算，設定感興趣區域以判斷長者或護理人員在畫面中進出與移動的方向。進而建立狀態移轉的觸發條件來偵測長者的寢居狀態是否安全。因此，一旦長者的狀態偵測為起身或下床時，系統將會發出警示來提醒護 理人員，以預防意外的發生。此外，透過Kinect感測器的夜間偵測的能力，讓系統在夜間沒有燈光的情況下也可以正常運作。同時，本系統也有記錄長者寢居狀態資訊的功能，可以讓護理人員輕易瞭解長者寢居時每個狀態的時間 點。透過一組測試影片的實驗中，結果顯示提出的方法高達 97.3%的狀態分析正確率。未來也可以進一步將這些狀態提供給家屬，以便於家屬瞭解其親人在長照中心或養護中心的狀況。

The demands of elder care and long-term care are increasing as the population of elder persons is increasing. However, nursing personnel have to take care of many elders in the night and in the elderly care center. A falling accident of an elder person may occur without notice by the personnel. It caused the problem of lawsuit or compensation. In order to solve the above problem, a system is developed in this project by utilizing Kinect depth sensor for analyzing the behavior of an elder in the bed. A state diagram is defined to model the behavior. When two states, sitting on the bed and getting out of bed, are detected, a siren is issued to remind the nursing personnel for preventing the possible accident. The Kinect depth sensor can be operated under very low brightness. It enables the system can monitor the behavior of an elder all the time. A set of testing videos with various scenarios is used to evaluate the performance. The experimental results show that the proposed method can analyze the behaviors of an elderly with 97.3% correction rate. In addition to the dangerous behavior, all the possible behavior can be monitored by the system for providing further services.