Fast and accurate primary user detection with machine learning techniques for cognitive radio networks
| Autor: | B. Sankaragomathi, G. A. Pethunachiyar |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Control and Optimization
Computer Networks and Communications Computer science Orthogonal frequency-division multiplexing Decision tree Machine learning computer.software_genre Multiplexing Spectrum management Naive Bayes classifier K-Nearest nighbor Electrical and Electronic Engineering Primary User Support vector machines business.industry Support vector machine Noise Cognitive radio Transmission (telecommunications) Hardware and Architecture Signal Processing Orthogonal Frequency Division Artificial intelligence business computer Information Systems |
| Popis: | Spectrum decision is an important and crucial task for the secondary user to avail the unlicensed spectrum for transmission. Managing the spectrum is an efficient one for spectrum sensing. Determining the primary user presence in the spectrum is an essential work for using the licensed spectrum of primary user. The information which lacks in managing the spectrum are the information about the primary user presence, accuracy in determining the existence of user in the spectrum, the cost for computation and difficult in finding the user in low signal-to noise ratio (SNR) values. The proposed system overcomes the above limitations. In the proposed system, the various techniques of machine learning like decision tree, support vector machines, naive bayes, ensemble based trees, nearest neighbour’s and logistic regression are used for testing the algorithm. As a first step, the spectrum sensing is done in two stages with Orthogonal Frequency Division Multiplexing and Energy Detection algorithm at the various values of SNR. The results generated from the above algorithm is used for database generation. Next, the different machine learning techniques are trained and compared for the results produced by different algorithms with the characteristics like speed, time taken for training and accuracy in prediction. The accuracy and finding the presence of the user in the spectrum at low SNR values are achieved by all the algorithms. The computation cost of the algorithm differs from each other. Among the tested techniques, k-nearest neighbour (KNN) algorithm produces the better performance in a minimized time. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|