CYCLOSTATIONARY SPECTRUM SENSING TECHNIQUES FOR COGNITIVE RADIO

Abstract

A cognitive radio by virtue of its ability to sense and adapt to the dynamic spectrum scenario, can increase the spectral efficiency. In order to be non-invasive, a cognitive radio must adhere to strict benchmarks in the quality of spectrum sensing for primary users of a band. Thus, spectrum sensing has a major role to play in cognitive radio. Many algorithms have been proposed to enable spectrum sensing such as energy detection and cyclostationary detection. IEEE 802.22, the first standard for cognitive radio devices, imposes strict requirements for the detection and false alarm probability on all spectrum sensing devices at SNR up to -20 dB. This requires use of robust spectrum sensing techniques. Energy detection is the simplest and near optimum technique that is widely used for spectrum sensing. However, its performance is drastically affected by uncertainty in noise variance due to SNR wall [1]. Cyclostationary detection can exploit spectral correlation present in most modulated signals to reliably detect signals even at low SNR. All general QAM signals exhibit distinct cyclic frequencies depending on their carrier frequency, baud rate etc. which can help to distinguish between the SOI (Signal of Interest) and interference. The filter structures for optimum MMSE estimation of cyclostationary signals are frequency shift (FRESH) filters. The theory of cyclic Wiener filtering theory, developed by Gardner [2], forms the basis of LCL (Linear Conjugate Linear) filtering used in FRESH filters. By adding appropriately frequency shifted versions of a cyclostationary signal, FRESH filters can provide significant gains for cyclostationary detection. Hence, it is intuitive to apply FRESH filters for spectrum sensing in the cognitive radio context. For upcoming wireless standards like Wi-MAX and LTE, OFDM (Orthogonal Frequency Division Multiplexing) is used because of the advantage of multicarrier transmission. Spectrum sensing for OFDM signals is especially challenging due to the cancellation of cyclostationary features and efficient detection algorithms for OFDM need to be developed. This thesis presents a comparative analysis between cyclostationary and energy detection and proposes FRESH filter based detection for cyclostationary spectrum sensing, with supporting simulation results. For spectrum sensing in OFDM, it develops an optimal Neyman-Pearson detector for induced cyclostationarity in CP-OFDM and shows that the proposed detector outperforms energy and cyclostationary detection techniques in the detection performance. iii