Parameter estimation of exponentially decayed sinusoidal signals based on high-order cumulant ESPRIT algorithm
Aiming at the problem that the actual environmental noise in engineering applications is mainly manifested as Gaussian colored noise and the algorithms for processing Gaussian white noise fail, a fourth-order cumulant ESPRIT algorithm is proposed for the estimation of the frequency and attenuation factor of multicomponent attenuated sinusoidal signals in Gaussian colored noise environments. First, the relationship between the fourth-order cumulants and the autocorrelation and intercorrelation matrices in the observed samples is derived to find their fourth-order cumulant matrices. Second, the generalized eigenvalue decomposition of the fourth-order cumulants is performed, and the signal attenuation factor and frequency estimates can be obtained based on the generalized eigenvalues. Finally, the proposed algorithm is validated by simulation experiments. The average estimation errors of the proposed algorithm for the angular frequency and the attenuation factor of the multicomponent fading sinusoidal signal are 0.0020π rad and 0.0020 at the hybrid signal-to-noise ratio of 0 dB. Compared with ESPRIT and Prony algorithms, the proposed algorithm has stronger noise suppression ability and higher parameter estimation accuracy in Gaussian white noise and Gaussian colored noise backgrounds.
exponential attenuation sinehigh order cumulantsESPRIT algorithmattenuation factor estimationfrequency estimation
万方数据
维普
