Hidden Markov Model-based Bandwidth Selectionfor Cognitive Network

• Main Authors: Yu,Ming-Yu, 余明逾
• Other Authors: Pao-Ann Hsiung
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/41930567477761115436

碩士 === 國立中正大學 === 通訊資訊數位學習碩士在職專班 === 101 === With the growing of wireless communication technology and mobile device, a lack of spectrum resource has become a serious problem. The concept of spectrum sharing in cognitive radio is regarded as a solution of spectrum usage rate. The concept of cognitive radio is to sense the temporal usable spectrums and then occupied the usable spectrums to transmit data in order to optimize the spectrum usage rate. However, the technologies of spectrum sensing do not always work well so that the spectrum usage rate is degraded. In this thesis, we proposed an HMM-based approach for spectrum sensing, that is, the state of spectrum occupation is represented by the HMM. However, the time labels are often necessary for model training and the time labels are labor cost and time consuming. Therefore, we employed the Kalman filter to detect the state boundaries instead of equally cut in initializing model. For the experiments, a simulative dataset is generated to evaluate our proposed approach. In implicit dataset that states implicitly change, the recognition performance can achieve 69.77% accuracy. In explicit dataset that states explicitly change, our approach finally can obtain 74.25% accuracy. Keyword – cognitive radio, spectrum sensing, hidden Markov model, Kalman filtering, spectrum management