Abstract
Recently, the patient-centric wireless medical environment has become more and more important. It not only provides patients’ physiological information, but also helps to save time on taking physical examinations in hospitals, which would further extend patients’ life span without increasing the medical cost. Therefore, we took advantage of Android smart phones and developed a ”Portable Healthcare System”. When emergencies occur, the patients can be traced by the devices and be rescued as soon as possible. In this thesis, we propose a novel indoor positioning method deploying WiFi access points (APs) called N-Cluster k-NN algorithm, which does not cost extra money for infrastructures and still offers decent accuracy comparing to other indoor positioning techniques. According to the offline simulation, the complexity of N-Cluster k-NN is low, while the accuracy can be up to 98.67%. In brief, the complexity and the calculation are greatly decreased, while the accuracy is still maintained. In chapter 3 and 4, we’ll talk about the Android-based platform and the Android application with intuitive user interface and quick access to changing parameters, in order to briefly demonstrate our result to determine the location of the users in real time. By this Android application, we’ll introduce the implementation of the online positioning. Online positioning was tested under N-Cluster k-NN with the optimal parameters obtained from offline simulations. The experimental environment was the Laboratory of Reliable Computing (LaRC) on the fourth floor in the TSMC building in National Tsing Hua University (NTHU). The whole environment was equally divided into twelve grids, which is about the size of a normal ward. With the background dimension in 227.59 m2, the average error rate of our algorithm is 4.75%, and the average error distance is 3.728 m. Compared to other designs, the accuracy of our algorithm does not differ much. In chapter 5, we further applied the algorithm above in National Taiwan University Hospital (NTUH). With the background dimension in 134.04 m2, the average error rate of our algorithm is 5.75%, and the average error distance is 3.328 m. Even in a different environment, the system is able to achieve a comparable result. Moreover, the complexity in NTUH is about a third less than it was in the LaRC. With such accuracy and portability, precise positions of every patient are sent to the cloud server and computed in real time, which enables doctors to be fully informed by their mobile handsets, with the minimum energy consumption and the longer duration the device can stand by.