A linear phase estimation technique for interferometric signals

In this paper we compare the performances of different estimation techniques applied in the case of the interferometric signal. Interferometric techniques have numerous applications, notably in SAR and SAS systems. The goal in interferometric systems is to estimate the instant of zero phase crossing which enables range/depth estimation. Current phase estimation techniques rely on the stationarity of the phase difference of two received signals. In this paper we assume a non-stationary phase difference, more precisely a linear function of time. This linear phase difference coupled with a classical complex Gaussian model is used to infer a statistical model of the received signals. The statistical signal model highlights the interpretation of phase delay estimation as a spectral estimator. More precisely, as the estimation of the angular frequency and initial phase of a single sinusoid. Using the proposed signal model several estimation techniques are tested. In estimating the signal phase two approaches can be considered: working directly with the phase by use of different regression techniques; or utilizing the complex interferometric signal. The first approach has the disadvantage of having to deal with 2kπ ambiguities while the second approach allows the use of different classical spectral estimators like the maximum likelihood (MLE) and AR models. We conclude by a performance analysis of these methods, in the case of the interferometric signal, which will show that depending on the SNR values the optimal estimator will differ. For SNR values grater than 2dB the MLE is shown to provide better performances while for lower SNR values a weighted linear regression technique (WLR) provides better results.