A Comparative Analysis of Several Back Propagation Algorithms in Wireless Channel for ANN-Based Mobile Radio Signal Detector
S. Roy Chatterjee1, R. Mandal2, M. Chakraborty3
1Prof. S. Roy Chatterjee, Department of Electronics and Communication Engineering, Netaji Subhas Engineering College, Kolkata, India.
2Mr. R. Mandal , Department of Electronics and Communication Engineering Netaji Subhas Engineering College, Kolkata, India.
3Dr. M. Chakraborty, Department of Information Technology, Institute of Engineering and Management, Kolkata, India.
Manuscript received on August 05, 2013. | Revised Manuscript received on August 11, 2013. | Manuscript published on August 15, 2013. | PP: 31-37 | Volume-1 Issue-9, August 2013. | Retrieval Number: I0414081913/2013©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Application of Artificial Neural Network (ANN) in cognitive radio has received considerable attention to incorporate artificial intelligence in cognitive radio based communication system. This paper introduces multilayer feed-forward neural network (MFNN) for spectrum sensing to detect the primary users. This in turn would enable the detector to identify the vacant bands that are devoid of primary users. As the accuracy of detection depends on the structure of the network and on the learning algorithms, an MFNN is trained with different back propagation algorithms varying the number of hidden neurons to find out the best suitable structure of the MFNN for spectrum sensing in different conditions of the wireless channel. Distinct cyclostationary features of different primary users are extracted to generate the input feature vectors for the MFNN as these features are well accepted for signal detection in low signal-to-noise ratio (SNR). The False Alarm Rate (FAR) of the detector is also evaluated with SNR and multipath delay. Simulation results prove that MFNN is suitable for designing a highly robust vacant band detector in the time varying wireless channel as it provides low and almost constant FAR in high multipath delay and low SNR RF environment.
Keywords: Back propagation algorithms, cognitive radio, false alarm rate, multilayer feed forward neural network, spectrum sensing.