A Novel Precolouring-Random Demodulator Architecture for Compressive Spectrum Estimation

Abstract

One of the main challenges of conventional spectrum estimation methods in cognitive radio applications is the very high sampling rates involved, which imposes significant operating demands upon the <i>analog-to-digital converter</i> (ADC). This has given impetus to employing <i>compressive sensing</i> (CS) techniques, such as the <i>random demodulator</i> (RD) structure to relax the input ADC specification. It has been recently shown the RD spectrum estimation performance for quadrature<i> phased shift keying</i> (PSK) modulated signals can be significantly improved in terms of <i>spectral concentration</i> and signal-to-noise ratio, when signals are precoloured by an <i>autoregressive</i> (AR) filter. This paper presents an extended AR-RD architecture, which provides enhanced CS capability for higher-order digital modulation schemes, including 16 <i>quadrature amplitude modulation</i> (16QAM), 64QAM and <i>binary</i> PSK (BPSK). Quantitative results corroborate the improved CS performance of the AR-RD structure for higher-order modulations schemes, which provides a propitious design trade-off between AR-RD complexity, latency and CS performance.