A Parallel Computing Based Spectrum Sensing Approach for Signal Detection under Conditions of Low SNR and Rayleigh Multipath Fading

Quickly and accurately detecting signals over a wide frequency range under conditions of low SNR and multipath fading is one of the most challenging requirements for dynamic spectrum access devices. The cyclostationary feature detection method is attractive for detecting primary users because of its ability to distinguish between modulated signals, interference, and noise at low SNRs. However, a key issue of cyclostationary signal analysis is the high computational cost arising from the large number of required complex convolution operations. This complexity increases in proportion to signal bandwidth if frequency resolution is held constant. To reduce the computation requirements, we use parallel computing in cyclostationary feature analysis running on a Cell Broadband Engine (Cell BE). Specifically, we parallelized the FFT accumulation method (FAM) algorithm to calculate spectral correlation functions (SCF) on four available Synergistic Processing Elements in a PlayStation 3. We analyze our algorithm's computational speedup on a Cell BE compared to a GPP based version. We analyze the impact of AWGN and Rayleigh multipath fading on signals' SCF features in the bi-frequency plane (with axes of cyclic-frequency and frequency). Next, we design a spectrum sensing algorithm which uses the distinct SCF pattern of each modulated signal to detect its existence. We run this algorithm on simulated signals including M-ary Phase-shift keying (PSK), Frequency-shift keying (FSK), Quadrature amplitude modulation (QAM), and PSK based Orthogonal frequency-division multiplexing (OFDM) under conditions of multipath fading and different levels of SNR. We give numerical results for the detector's performance.