CZT-Based High Performance Frequency Estimation

Accurate and high-performance frequency estimation of a noisy sampled sinusoid sequence with short time length and adjacent overlapping has been attracting top focus in the field of signal processing. Based on the CZT of a single-tone sequence, this paper presents a novel improved Rife CZT-interpolated frequency estimator, which develops an improved Rife algorithm on the symmetry of CZT Linear Spectrum Model by the magnitude relationship between peak and its adjacent CZT coefficients. Computer Monte -Carlo simulations show that the algorithm is effective and the proposed estimator outperforms the traditional Rife-Jane's estimator in accuracy, the CRLB-approximation RMSE of estimation and discriminating capability for adjacent overlapping. It is also shown that the CZT-based estimator has as lower SNR threshold as to about -12 dB and good estimation performance even in the case of very short samples, which is quite suitable for applications of strict stationary requirement.