一個使用環場攝影機及多重辨識器的跌倒偵測系統

由於生活水平的提高與醫療技術的進步，醫療照護機構也越來越完善，使得人類的平均壽命得以延長，老年人口也急速增加，因此對於老人的看護也越來越受到重視。對於老人的看護中，跌倒意外的發生已經成為看護的重點之一，因為跌倒意外的發生會對老人的生命健康造成極大的威脅。本文旨在發展一套適用於療養照護機構中並且針對老人或者是住院病患的跌倒偵測系統，期望在最快的時間內偵測出跌倒事件，並且通知最近的醫療照護人員前往看護。本系統採用具有360度視角的環場攝影機，有別於一般跌倒偵測系統必須使用多台有限視角的傳統攝影機來監控全場，而造成較高的系統成本。對於跌倒辨識方面，在受測者有不同的跌倒方向時，本論文亦提出不同的特徵，並且搭配多級辨識器以達到較高的系統效能。由實驗得知，當受測者的行走路徑有較大變化時，本系統也不容易因此而有誤判的情況發生，且當受測者在不同方向發生跌倒時仍有很好的辨識效能。由實驗結果得知，本系統的辨識率可以到達0.87，系統有效性亦到達0.88。如能加入其他生醫訊號分析系統例如脈搏訊號、ECG (Electrocardiogram)或加速規等，可以使系統具有更高的辨識效能，降低誤判。

關鍵字

跌倒偵測；多級辨識器；環場攝影機

並列摘要

With the advancement of medicine and living standard and better health-care institutions, the life span of people is longer and the number of the elderly increases rapidly. Consequently the health care of the elderly attracts more and more attention. The fall accident has become one of the most important issues in the health care because it can present a great threat and injury to the elderly. This thesis proposes a fall detection system for the elderly and the patient, hoping to detect the fall accident immediately when it happens and notify the medical personnel to provide the emergency care. In our system we use an omni-directional camera that has a 360∘viewing angle. Unlike other fall detection systems, where multiple traditional cameras are used to capture the whole scene of the surveillance area, the proposed system operates simpler and costs less. In fall detection, we use different recognition features to allow for different falling directions. Furthermore, a two-stage multiple classifier is adopted to enhance the detection performance of our system. The experiment results show that even when walking paths have changed significantly our system will not give a false alarm easily, and our system can also perform well when various falling directions occur. With the proposed system, the fall down recognition rate and the specificity can reach 0.87 and 0.88, respectively. Furthermore, the false alarm chance is slim in the testing environment where a person walks randomly with variously behaviors or movements. If we can obtain and analyze other medical signals, such as the pulse signal, ECG, or accelerometer as well, the system performance can be enhanced and the false alarm rate can be reduced further.

並列關鍵字

Fall detection ； omni-directional camera (MapCam) ； multiple classifiers

參考文獻

[1]林茂榮以及王夷暐，社區老人跌倒的危險因子與預防，台灣公共衛生雜誌第23卷4期，民國九十三年。

[26]張勝仁，一個可增強車牌字元辨識效能的多重辨識器設計，中原大學電子工程研究所碩士論文，民國91年。

[5]I. Haritaoglu, D. Harwood, and L. Davis, “W4: real-time surveillance of people and their activities,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22, no. 8, pp. 809-830, Aug. 2000.

[6]Y. Cui and J. Weng, “Hand segmentation using learning-based prediction and verification for hand sign recognition,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, Puerto Rico, pp. 88-93, June 1996.

[7]A. Bobick and J. Davis, “Real-time recognition of activity using temporal templates,” in Proc. of IEEE Workshop on Appication. of Computer Vision, Sarasota, Florida, pp. 39-42, Dec. 1996.

國際替代計量

一個使用環場攝影機及多重辨識器的跌倒偵測系統

全文下載

主題瀏覽