Indoor Positioning Using Modified Bluetooth Fingerprint Method

• Main Authors: Ho, Hao, 何浩
• Other Authors: Lin, Shiou-Gwo
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/56989018538229848743

碩士 === 國立臺灣海洋大學 === 通訊與導航工程學系 === 104 === Positioning systems have played a major role in people’s lives since the Global Positioning System (GPS) became publicly available. In nowadays, almost everyone has a device with positioning capabilities. However, GPS will cause high distance error rate in indoor environment. Due to development of wireless communication, more and more researchers have focused on developing indoor positioning system with Received Signal Strength Indicator (RSSI). Bluetooth Low Energy (BLE) devices are inexpensive, small, long battery life and do not require an external energy source. This thesis use fingerprint-based localization systems based BLE devices. Fingerprint-based localization systems does not rely on calculating signal fading. It is usually divided into two main phases, on-line phase and off-line phase. Collect the RSSI patterns in fields where localization is needed into a database which called radio map. During the on-line phase, collect current RSSI pattern and compare it with the radio map to identify its possible location. Based on RSSI, a well-known method called K-Nearest Neighbors (KNN) has widely been used. It calculates the Euclidean distance on RSSI space between RSSI of the training points. Chose k training points that have smallest Euclidean distances to estimate the current position. As a widely applied clustering method, K-means has the merits of fast running and moderate clustering quality. To improve the method, this thesis replace KNN with K-means to cluster a set of training points with their Euclidean distances and coordinates. Choose the smallest average Euclidean distance to be the target cluster to estimate position. As the result shows that K-means is more stable than KNN.