- 學位論文
一個使用移動歷史紅外線影像為基礎的睡眠動作偵測系統
A Sleep Motion Detection System Based on Infrared MHI Images
摘要
睡眠佔據人的一生約三分之一的時間。當夜間的睡眠品質不好時，會間接影響到日間的作息，進而使心理及生理受到損害。近年來，國內也逐漸重視睡眠品質，因而成立了許多睡眠中心，其中睡眠品質的偵測為工作重點之一，需要投入足夠的人力及精密昂貴的醫療器材，才得以維護其醫療與照護的品質。本論文提出一套以視覺為基礎，使用近紅外線攝影機且能應用於居家與醫療照護機構的睡眠行為偵測系統，希望此低成本系統能幫助醫生做後端的資料分析，迅速判斷出患者的睡眠狀況，以改善睡眠品質。 本系統分析移動歷史影像的向量，得到翻身特徵，接著透過翻身偵測及配合睡眠姿勢狀態圖的使用，進而得知當下的睡眠姿勢。另外，使用歷史移動影像的移動能量以及適合的門檻值，使得系統可以正確捕捉到一些微小動作的持續時間，提升整體睡眠品質偵測效率。 實驗結果顯示，本系統成功的解決了連續影像相減法的雜訊生成及特徵不明顯等問題，也成功偵測出睡眠姿勢。而在睡眠品質分析方面，系統分析結果頗接近實際的睡眠品質數值，因此可以證明本論文提出一個能有效偵測睡眠品質的分析系統。
並列摘要
One third of our life is spent in sleeping. If our sleeping quality is poor at night, it will indirectly affect our daily activities and harm our physical and psychological health. In recent years, domestic people are gradually paying attention to sleeping quality, thus a number of sleep centers are established, where monitoring sleeping quality is one of the major tasks. It requires sufficient manpower and sophisticated, expensive medical equipments to maintain certain levels of medical and health-care quality. This thesis proposes a vision-based sleeping behavior detection system using an infrared camera. The proposed low cost system can be applied to both home and health-care institutions, hoping that it can assist doctors to do data analysis and quickly determining and improving one’s sleeping quality. The proposed system analyzes vectors of the Motion History Imaging (MHI) to get the feature of the turn of body. Then, the system uses body turn detection and a sleeping status state diagram to get the current sleep posture. In addition, with the motion energy of the MHI and a suitable threshold, the system can accurately capture the duration of some small movements and enhance the detection efficiency of overall sleeping quality. Experimental results show that the proposed system has overcome the problem of noise creation and unobvious features of the Temporal Differencing approach for adjacent frames and has successfully detected sleep postures correctly. Finally, the result of the sleep quality analysis is close to the actual value, showing that we have proposed an effective sleep quality analyzing system.