Research On Methods And Implementation Of System For Collaborative Fall-detection Based On Confidence Model

With the advent of an aging era, more and more old people is facing the situation of living alone, the situation of “empty nester” is becoming much more serious. The elderly lacking for their family’s care, the family of them are becoming increasingly desirable for the smart device in the home. They hope to use the smart devices to serve the elderly. These devices could monitor the behavior of an older person living in the home, and so abnormal behavior could be detected so as not to lead to irreparable economic and moral damage. Nowadays, many researches using a single device to carry out fall detection has been studied, but the device they use are scattered and isolated, and are not interconnected. In view of this situation, we propose a method that multiple devices work together to perform fall detection, aiming at increasing the accuracy rate of detecting abnormal behavior i.e. fall, so as to serve the elderly users in the complex home environment. We have established a collaborative monitoring platform, monitor users can take consideration of the special needs of care staff to select the appropriate services mode for the elderly home user, in this way, we collaborate multiple devices to carry out fall detection, so that the elderly users in the home could be better served.In the first place, this paper analyzes the special needs of the family fall detection in complex home environments, explains the demand for devices with regard to fall detection and elaborates the method of fall detection and data fusion. Therefore, we proposed the problem that much importance should be attached to the cooperative “fall detection” service platform. Conducting collaborative “fall detection” of multiple devices, we need to consider the related technology of data communications and data fusion. And this paper describes the collaborative “fall detection” service targets.In the next place, the principle of the fall detection method and means to achieve the principles are described, and we studied data fusion methods. Confidence model has been proposed, and on this basis of this collaborative fall detection using data fusion technology is carried out. The specific related method is as follows: a method of fall detection based on threshold using phone; fall detection method based on support vector machine using Microsoft Kinect device, collaborative fall detection method based on the rules of logic and cooperative fall detection method based on D-S evidence theory.Furthermore, for the method for fall detection and data fusion method which are previously proposed, in this part, collaborative monitoring platform fall detection, a subsystem of fall detection based on threshold using phones and a subsystem of fall detection method based on support vector machine using Microsoft Kinect device have been designed and implemented.Last but not least, in order to verify the effectiveness and rationality of fall detection subsystems and cooperative fall detection monitoring platform, monitoring platform and function of fall detection subsystem was tested, and designed a fall detection testing scheme. According to the experimental scheme, fall detection subsystem performance and the performance of collaborative fall detection was analyzed and compared. This paper focused on comparing the accuracy rate and false alarm rate of single device fall detection method and collaborative fall detection method.