Improved, Approximate, Time-Domain ML Estimators of Chirp Signal Parameters and Their Performance Analysis

We derive an improved, approximate, maximum likelihood (ML) estimator for the parameters of a single-component chirp signal in the time domain, which has much lower computational complexity than that of the traditional estimators in the frequency domain. We term it the zero-order amplitude weighted phase-based estimator (AWPE). It is a weighted linear combination of the phases of the received signal samples, where the weights are determined by the amplitudes of the received signals. It achieves better performance than the only existing time-domain estimator in the literature that only exploits the phase information. A phase unwrapping (PU) algorithm is needed to recover the phase information correctly from the received signal samples. Several PU algorithms are compared and analyzed. A new, robust PU algorithm based on the first-order phase differences of the consecutive received signal samples is proposed. This new unwrapping algorithm has smaller PU failure probability than that based directly on the phases of the received signal samples. The need for PU can even be removed by using the second-order phase differences of the consecutive received signal samples. Two additional structures for the ML-based AWPE using the phase differences are derived starting from using the zero-order AWPE obtained. The first is based on the first-order phase differences and the second on the second-order phase differences. When the PU algorithm works perfectly, i.e., no phase unwrapping failure occurs, the estimators based on phase differences do not perform as well as the original one based on absolute phase information due to the increased noise variance. This is also applicable when the chirp parameters take on very small values. However, when no knowledge of the ranges of the chirp parameters is available, and the possible PU failures are taken into account, the estimators based on the phase differences could perform better. Performance analysis of the estimators is given based on an improved phase noise model. It is easy to verify that the estimates obtained are unbiased. The cramer-rao lower bound (CRLB) on the mean-square error (MSE) is derived. The MSE performance of the estimators approaches the CRLB at high signal-to-noise ratios. The variance performance of the estimators based on phase differences is analyzed, whose validity is demonstrated by simulation results.