With the advancement of the Internet of Things(IoT) and pervasive computing applications, it provides a better opportunity to understand the behavior of the aging population. However, in a nursing home scenario, common sensors and techniques used to track an elderly living alone are not suitable. In this paper, we design a location-based tracking system for a four-story nursing home - The Salvation Army, Peacehaven Nursing Home in Singapore. The main challenge here is to identify the group activity among the nursing home's residents and to detect if they have any deviated activity behavior. We propose a location-based deviated activity behavior detection system to detect deviated activity behavior by leveraging data fusion technique. In order to compute the features for data fusion, an adaptive method is applied for extracting the group and individual activity time and generate daily hybrid norm for each of the residents. Next, deviated activity behavior detection is executed by considering the difference between daily norm patterns and daily input data for each resident. Lastly, the deviated activity behavior among the residents are classified using a rule-based classification approach. Through the implementation, there are 44.4% of the residents do not have deviated activity behavior , while 37% residents involved in one deviated activity behavior and 18.6% residents have two or more deviated activity behaviors.

翻译：随着物联网和普适计算应用的进步，这为理解老年人口的行为提供了更好的机会。然而，在养老院场景中，用于追踪独居老人的常用传感器和技术并不适用。本文设计了一套基于位置追踪的系统，应用于新加坡救世军平安养老院这一四层楼的养老机构。主要挑战在于识别养老院居民的群体活动，并检测他们是否存在任何行为偏离。我们提出了一种基于位置的行为偏离检测系统，通过利用数据融合技术来检测偏离行为。为了计算用于数据融合的特征，采用自适应方法提取群体和个体的活动时间，并为每位居民生成每日混合规范。随后，通过比较每位居民的每日规范模式与每日输入数据之间的差异，执行偏离行为检测。最后，使用基于规则的分类方法对居民中的偏离行为进行分类。实施结果显示，44.4%的居民没有偏离行为，37%的居民涉及一种偏离行为，而18.6%的居民有两种或更多偏离行为。