Center Frequency Estimation for Non-gaussian Data Using a Non-linear Method, the Symmiktos Method

When estimating the Doppler frequency, it is common to use a covariance approach. This is especially true for applications such as underwater current profiling . The early developers of this technology made the assumption that the backscattered signal obeyed Gaussian properties. Based on that, a covariance approach have been used for the estimation of the center frequency I the Doppler signal. Due to a complex Doppler signal, the backscattering process is not always Gauusian, and the covariance method will create bias and wrong estimates. The Symmiktos Method have been developed to handle the estimation of the Doppler frequency in a more robust sense, and independent of wether or not the backscattering signal obeys Gaussian properties. The method is non-linear, and is performed in frequency domain. The Symmiktos Method also gives less bias; a problem that is inherent in the classical method, and one that is increased if the variance is decreased. Modeled data as well as real-life data is used to show typical behavior of the two methods. The analysis of the data clearly shows that the covariance method is more sensitive and biased in the presence of signals that do not obey the Gaussian assumption. Also, the large database with backscattering signals from four different locations has been used in the comparison of the two methods. The comparison shows a large difference between the two estimators. All tests on simulated signals have shown that the Symmiktos Method is superior compared to the covariance method for all the data analyzed.