Design and implementation of a distributed fall detection system based on wireless sensor networks

Xiaomu Luo 0 Tong Liu 0 Jun Liu 0 Xuemei Guo 0 Guoli Wang 0 0 School of Information Science and Technology, Sun Yat-sen University , Guangzhou 510006, China Pervasive healthcare is one of the most important applications of the Internet of Things (IoT). As part of the IoT, the wireless sensor networks (WSNs) are responsible for sensing the abnormal behavior of the elderly or patients. In this article, we design and implement a fall detection system called SensFall. With the resource restricted sensor nodes, it is vital to find an efficient feature to describe the scene. Based on the optical flow analysis, it can be observed that the thermal energy variation of each sub-region of the monitored region is a salient spatio-temporal feature that characterizes the fall. The main contribution of this study is to develop a feature-specific sensing system to capture this feature so as to detect the occurrence of a fall. In our system, the three-dimensional (3D) object space is segmented into some distinct discrete sampling cells, and pyroelectric infrared (PIR) sensors are employed to detect the variance of the thermal flux within these cells. The hierarchical classifier (two-layer HMMs) is proposed to model the time-varying PIR signal and classify different human activities. We use self-developed PIR sensor nodes mounted on the ceiling and construct a WSN based on ZigBee (802.15.4) protocol. We conduct experiments in a real office environment. The volunteers simulate several kinds of activities including falling, sitting down, standing up from a chair, walking, and jogging. Encouraging experimental results confirm the efficacy of our system. - healthcare [3]. Although falls are specific cases of healthcare, there is a significant research effort focusing on fall detection. This is due to the fact that accidental falls are among the leading causes of death over 65 [4]. According to the report in Chan et al. [5], approximately one-third of the 75 years or older people have suffered a fall each year. The fall of the elderly is a serious problem in an aging society [6]. The immediate treatment of the injured people by the fall is very critical, because it will not only increase the independent living ability of the elderly and the patient, but also release the pressure of the shortage of nurses. Therefore, how to design a rapid alarm system for fall detection has always been an active research topic on the elderly healthcare. Camera-based methods may realize fall detection for elderly people in a non-intrusive fashion. For example, Williams et al. [7] extracted the human target with the simple background subtraction method from the video, and then used the aspect ratio of the image of the body as the cue to determine whether the fall event happened. If the aspect ratio, i.e., the width of the person divided by height, is below a particular threshold, then we can assume that the person is upright; otherwise the person is assumed to have fallen. Rougier et al. [8] integrated the motion history image (MHI) and the variance of body shape information as the feature for fall recognition. Although there are so many works that demonstrate their efficiency [6], these studies are based on the assumption that the lighting conditions remain fairly stable. However, this assumption does not always hold in everyday life. The camera-based analysis may be influenced by the change of illumination and the shadow, and accurate body extraction from video is still a thorny issue in the computer vision community. With resource constrained sensor nodes, sophisticated algorithms are not preferable choices. In addition, the camera-based method will infringe privacy; no one likes the feeling of being monitored by a camera all day long. Is it possible to find another sensing method to detect the fall? This is the motivation of our research. In the WSN-portion of the IoT, the choice of sensing modality is critical. Recently, there has been a growing tendency to research sensing modalities with pyroelectric infrared (PIR) sensors [9-11]. The PIR sensor is a kind of thermal imaging technologies and responses only to temperature changes caused by human motion. It has the promising advantages to overcome the limitations of the traditional camera-based sensing method, since the human motion information is acquired directly, without sensing redundant background and chromatic information. The output of the PIR sensors is low-dimensional temporal data stream, which avoids the high-dimensional data processing. However, PIR sensors provide fairly crude data from which it is difficult to acquire the spatial information. Thus, the primary goal of the sensing system design is to enhance the spatial awareness of the PIR sensors, and to capture the spatiotemporal feature of the fall. In this article, we design and implement a system, SensFall, which can detect the fall efficiently and efficaciously. Sensing model design is the most important process of our (...truncated)