Spectrum sensing and identification in cognitive radio using independent component analysis -

Abstract

Wireless networks are mainly based on fixed spectrum assignment policies. However, due to the increase in the penetration of communication systems and the congestion of spectrum allocation, the FCC decided to transfer to a new paradigm that allows wireless technologies and end users to share spectrum; this concept is referred to as cognitive radio. A cognitive radio system can smartly sense spectrum changes and adapt various radio frequency parameters in accordance with the changing environment to maintain an efficient and reliable communications. The two main challenges for cognitive radios are that it should not interfere with the licensed users and should vacate the band when required. Thus, there is a need for fast, efficient and robust spectrum detection schemes. However, every spectrum detection approach has its limitations as well as its advantages; these limitations can vary from implementation issues to environmental issues. In order to overcome the limitations of the detection algorithms, blind source separation algorithms have been proposed to improve reliability. Blind spectrum sensing require limited information of the source signals and the channel characteristics. The thesis main objective is to implement and evaluate a blind spectrum detection method using a hybrid Independent Component Analysis (ICA) - Energy Detection (ED) approach. The proposed hybrid approach is compared to spectrum sensing detection methods from the literature in terms of computational complexity and performance accuracy. In order to study realistic wireless scenarios, we perform simulations based on samples of signals that resemble existing technologies such as GSM (based on GMSK modulation) and LTE (based on OFDM modulation).