Artificial Neural Networks, Support Vector Machine and Energy Detection for Spectrum Sensing based on Real Signals

A Cognitive Radio (CR) is an intelligent wireless communication system, which is able to improve the utilization of the spectral environment. Spectrum sensing (SS) is one of the most important phases in the cognitive radio cycle, this operation consists in detecting signals presence in a particular...

Full description

Saved in:
Bibliographic Details
Published inInternational journal of communication networks and information security Vol. 11; no. 1; pp. 52 - 60
Main Authors Saber, Mohammed, El Rharras, Abdessamad, Saadane, Rachid, Aroussi, Hatim Kharraz, Wahbi, Mohammed
Format Journal Article
LanguageEnglish
Published Kohat Kohat University of Science and Technology (KUST) 01.04.2019
Subjects
Algorithms
Artificial intelligence
Artificial neural networks
Cognitive radio
Detection
False alarms
Frequencies
Keying
Neural networks
Performance evaluation
Power consumption
Radio signals
Signal detection
Signal transduction
Software
Spectrum allocation
Support vector machines
Wireless communication systems
Wireless communications
Online AccessGet full text

Cover

More Information
Summary:A Cognitive Radio (CR) is an intelligent wireless communication system, which is able to improve the utilization of the spectral environment. Spectrum sensing (SS) is one of the most important phases in the cognitive radio cycle, this operation consists in detecting signals presence in a particular frequency band. In order to detect primary user (PU) existence, this paper proposes a low cost and low power consumption spectrum sensing implementation. Our proposed platform is tested based on real world signals. Those signals are generated by a Raspberry Pi card and a 433 MHz Wireless transmitter (ASK (Amplitude-Shift Keying) and FSK (Frequency-Shift Keying) modulation type). RTL-SDR dongle is used as a reception interface. In this work, we compare the performance of three methods for SS operation: The energy detection technique, the Artificial neural network (ANN) and the support vector machine (SVM). So, the received data could be classified as a PU or not (noise) by the ED method, and by training and testing on a proposed ANN and SVM classification model. The proposed algorithms are implemented under MATLAB software. In order to determine the best architecture, in the case of ANN, two different training algorithms are compared. Furthermore, we have investigated the effect of several SVM functions. The main objective is to find out the best method for signal detection between the three methods. The performance evaluation of our proposed system is the probability of detection (Pd) and the false alarm probability (Pfa). This Comparative work has shown that the SS operation by SVM can be more accurate than ANN and ED.
ISSN:2073-607X
2076-0930