Signal detection and automatic modulation classification based spectrum sensing using PCA-ANN with real word signals

Abstract

Cognitive radio has been proposed as an optimal solution to increase spectrum efficiency, by exploiting unused portions of radio spectrum. The first and the most important function in cognitive radio equipment, is spectrum sensing, wherein a cognitive user must sense his environment to detect a free band, then use it temporarily without causing any interference to the primary user (PU). It has been shown that the energy detection is the most chosen method, to detect the spectrum holes in the case where there is no a priori information about the PU. In this paper, we are describing tow stages architecture: The first stage is used to detect the presence of the PU with an improved energy detection technique, and the second is utilized to classify the modulation type, by combining principal component analysis to extract signal features and neural network to make the decision. Performance evaluation of our proposed architecture has been done by real word signals. 7960 A. Elrharras et al.

Artificial Neural Networks, Support Vector Machine And Energy Detection For Spectrum Sensing Based On Real Signals

“…Substituting ( 17) and ( 18) into Equation ( 16), the optimal separating hyperplane can be obtained by solving the following dual representation of the optimization problem: 𝑦 𝑖 = 0 , 𝛼 𝑖 > 0 By solving this dual Lagrange function (19), 𝛼 is evaluated. Consequently, 𝜔 is evaluated out from (17), and 𝑏 can be easily calculated from (20):…”

“…If the signal energy exceeds the threshold, we declare the presence of the PU, otherwise it is absent. During the last years soft computing techniques like artificial neural networks (ANN) and support vector machine (SVM), have become extremely successful discriminative approaches to pattern classification [14][15][16][17][18][19][20]. In our context, we propose an implementation of ANN and SVM for SS operation to detect the PU signal; we focus on different ANN training algorithms and SVM functions that can be applied on the set of input data patterns.…”