Improved Fast Cyclic Spectral Estimation Algorithm and Performance Analysis

An improved fast algorithm is proposed for cyclic spectral estimation, which decreases the requirement of data quantity, without reducing the performance. The windowed overlapped data processing, which is used in time smooth method, is introduced into the frequency smooth algorithm to cut down the original cyclic spectral estimation variance and improve the estimation quality. The paper deduces the estimators asymptotically expressions of mean, variance, resolution and computational complexity. The theoretical analysis and simulation results prove that the improved algorithm displays better performance than DFSM under the same conditions. The new method is an efficient estimator of cyclic spectrum in the low SNR, high resolution demanding and small data quantity environment.