Research On Key Techniques Of Dynamic Spectrum Cognitive Wireless Communication

The key techniques of cognitive radio can be summarized in three aspects: spectrum sensing, spectrum sharing and spectrum management. In this paper, spectrum data compression, spectrum data mining, chaotic sequence prediction and dynamic spectrum cooperative cognitive wireless communication system architecture are investigated. The innovative work can be divided into the following four parts:1. For solving the problem of spectrum data transmission which consumes too many resources, this paper proposes a spectrum data compression algorithm. It is able to remove redundant information of noise and reduces the amount of interactive data. By analyzing the characteristics of the spectrum data and the deficiencies of traditional data compression techniques, the algorithm divides spectrum data into signal part and noise part. Different part is compressed by different method. Simulation results show the algorithm obtains improvement of compression. Furthermore, we find the neighborhood similarity of spectrum through analyzing the spectrum property. A segmental compression algorithm is proposed, it utilizes the neighborhood similarity to improve the energy focusing of DCT coefficients. The compression simulation shows the segmental compression algorithm improves compression performance in most cases.2. By analyzing spectrum data characteristics, a spectrum data mining method base on increment operation is proposed to reduce the large computation of spectrum data mining. System fuses new information and existing data base by increment operation, and does not need global mining. The method effectively reduces the amount of computation. For multi-dimensional and sparse, non-continuous and variability of the spectrum data, a lot of mining information is analyzed. Besides these, this paper utilizes prediction technique to generate a reliable frequency map which is used to assist frequency selection.3. Base on least square support vector machine (LSSVM), three chaotic time series prediction algorithms under different scenarios are investigated. First, a prediction algorithm that consists of an iterative error correction and LSSVM is proposed. The algorithm creates multiple predictive models via the method of iterative error correction to approximate the chaotic system mapping function and obtain significant improvements of predictive performance. Then, a local LSSVM prediction algorithm of small scale network traffic base on correlation analysis is proposed. It optimizes the prediction model training set by correlation analysis and removes the noise components from training data. Simulation shows the algorithm improves the prediction accuracy and reduces the amount of computation. Finally, a power load LSSVM prediction algorithm base on K-means classification is proposed, and the algorithm obtain improvement of multi-step prediction performance.4. The wireless communication system architecture with dynamic spectrum cognitive is designed. And it can be used in satellite communications. Data mining with mass spectrum data is done in the control center of system and the regularity information of spectrum environment can be obtained. The satellite of system collects real-time spectrum sensing data. The system improves the intelligence of frequency selection while taking advantage of regularity information and real-time spectrum sensing data.