| Summary: | 碩士 === 元智大學 === 資訊工程學系 === 96 === The next generation wireless networks integrate several network technologies to provide an intelligent and ubiquitous communication. The success of next generation
mobile networks relies on the degree of seamless mobility support among heterogeneous technologies. In the past, two mobile-controlled handoff mechanisms, Network Discovery with User Motion Detection (NDMD) and Momentum of Received Signal Strength (MRSS) have been proposed to trigger handoff procedure earlier to reduce handoff delay. NDMD and MRSS use different smooth factors to calculate Exponential Weighted Moving Averages (EMA) of Received Signal Strength (RSS) to determine the motion of the MN. However, both of NDMD and MRSS have not mention a general analytic framework of the motion prediction. Moreover, the adjusting mechanism of different smoothing factor is not provided. This research
proposes a general analytic framework and a Multi-Resolution based Motion Detection algorithm (MRMD). In the framework, the smoothing factor assignment and RSS analysis are modeled by the Discrete Wavelet Transform (DWT) theory. Additional, a probability neural network based learning model is applied to learn the adjusted smoothing factors that trained by the movement of mobile nodes. Based on the multi-resolution theory and the probability neural network, the proposed MRMD
algorithm can detect user motion efficiently. In the simulation, we compare the performance of MRMD with NDMD, MRSS and traditional Dwell time based handoff approach. The simulation results confirm that MRMD has the highest motion
detection success rate, the lowest power consumption rate, lowest number of failure handoff and lowest number of connection loss.
|
|---|
