Research On Spectrum Prediction And Spectrum Sensing Technology Based On Hidden Markov Model

With the development of communication technology, cognitive radio has become a popular research direction of this field. The main idea is to allow the radio to master learning ability. It should learn to interact information with the surrounding environment and sense the spectrum "holes". At the same time, the secondary users should reduce the interference to the primary users. Among them, spectrum sensing is the key to the whole system. High performance spectrum sensing method should avoid the collision probability between primary users and secondary users. So, it can reduce the interference and improve the communication quality. Furthermore, in order to make full use of the idle spectrum to communicate, secondary users should detect the spectrum "holes" rapidly. Accurate and rapid spectrum sensing can avoid the interference and improve the throughput of the whole system. So, how to achieve rapid and accurate spectrum sensing is the main subject of this thesis.Aimed at rapid and accurate spectrum sensing, predicting before spectrum sensing is necessary. Secondary users predicts the channel by learning spectrum sensing historical data. And sensing the most possibility spectrum "holes" firstly based on the prediction results. An adaptive cooperative spectrum prediction method based on discrete hidden Markov models is proposed to solve this problem. In this method, the secondary users obtain the channel state training set according to the spectrum sensing history information. The channel status in the past M slots can be used as testing matrix to match with the training set, then predicts the channel state according to matching similarity degree on next slot. The simulation results show that the proposed method can improve spectrum prediction accuracy obviously.Then, in order to make full use of spectrum sensing history information and overcome the information distortion problem that caused by discrete hidden Markov models in the process of vector quantization, a spectrum sensing method based on continuous hidden Markov models is proposed in this thesis. Research has shown that the collected energy values when channel occupied status and channel idle status meet different Gaussian distribution. So, spectrum sensing method based on pattern recognition with continuous hidden Markov models can be used. Judging the channels is whether occupied or idle based on the match result. The simulation results show that this method has higher accuracy in spectrum sensing.