Blind Estimation of Carrier Frequency and Symbol Rate Based on Cyclic Spectrum Density

Single channel steepest descent algorithm for the correction of cycle frequency error

“…5 a is 500. Comparisons are carried out against existing methods based on TA‐MSE [1], cyclic autocorrelation (CA) [2], CSD [3], cyclic domain energy profile (CDEP) [4] and robust cyclic autocorrelation (RCA) [5]. The step size k is set at 5, δ = 2 × 10 −4 ; J TA ‐MSE , the number of points used to calculate TA‐MSE is 2; iteration times for calculating RCA is 10; and the number of signal frames concerning the calculation of CDEP is 2.…”

“…Most man‐made signals prefer a cyclostationary (CS) model rather than a stationary one, or equivalently, they exhibit spectral correlation , that is correlation between time‐shifted and frequency‐shifted versions of themselves, with the frequency shifts related to their hidden periodicities [1]. In the last decade, beneficial characteristics of signal cyclostationarity have received much attention, and have been intensively exploited in varied domains including signal parameter estimation [2–5], interference suppression or signal separation[1, 6–8], array beamforming [9–12] and recently, spectrum sensing [13–16].…”

“…In signal parameter estimation, cyclic feature functions including cyclic autocorrelation function [2], cyclic spectrum density (CSD) [3] is exploited to estimate the carrier frequency and the symbol rate of digitally modulated signals in non‐cooperative communications or other blind scenarios, for purposes of modulation classification, demodulation and so on. In interference suppression or signal separation, frequency‐shift (FRESH) filters are constructed to suppress narrowband interference (NBI) [1, 7] or extract the signal of interest (SOI) [6], and the accurate information of cycle frequencies is required for the filter to function.…”

“…In [2], the estimation of CF is achieved via searching the maximum of the amplitude of the cyclic autocorrelation. In [3], a blind estimation method based on CSD is proposed to estimate CFs with respect to non‐cooperative communication. In [4], the authors point out that estimators based on the maximisation of cyclic autocorrelation function (such as the one in [2]) are susceptible to the pulse shape variation and multipath fading, and a more robust estimator adopting the cyclic domain energy profile (CDEP) as the feature is proposed therein.…”

Chip rate and pseudo‐noise sequence estimation for direct sequence spread spectrum signals

“…Many techniques have been proposed to detect the chip rate of DSSS signals, such as the algorithm based on cyclic spectrum density (CSD) [3][4][5], the algorithm on wavelet transform (WT) [6][7][8] and the delay-and-multiply (DM)-based algorithm [9,10]. In the CSDbased algorithm, the cyclostationary statistics are exploited to estimate the chip rate by the peak value in the cyclic domain [3][4][5]. In the environment of low signal-to-noise ratio (SNR) or insufficient data, the performance degradation of CSD-based algorithm occurs, and this algorithm usually needs more data samples to improve its performance.…”