Improving the accuracy of parameter estimation for real exponentially damped sinusoids in noise

The mean filter of the forward predictor (FP) and backward predictor (BP) is proposed as an algorithm giving improved accuracy of parameter estimation for real exponentially damped sinusoids in noise. In this algorithm the mean values of the FP estimate and the BP estimate for frequencies and damping factors of multiple real exponentially damped sinusoids are calculated instead of the BP estimate. Monte Carlo simulations show that the estimate of large damping factors is unbiased over a wide range of the normalised frequency. The standard deviation of the estimate for large damping factors by the mean filter of forward and backward predictor (MFB algorithm) is slightly larger than that by the improved KT algorithm.