Statistical estimation of the parameters of a moving source from array data (Q1083167)

For the model \(X_ 1(t)=S(t)+e_ 1(t)\); \(X_ 2(t)=S(a_ 0+b_ 0t)+e_ 2(t)\), with known 'shifted signal', given by \(S(a+bt)\), a,b parameters and e(t) noise, the problem is to estimate the variable time delay, i.e. (a,b). One determines \((\hat a,\hat b)\) which maximizes the criterion: \[ Q_ T(a,b)=\sum_{| s| <T/2}\psi (\omega_ s)\underline d^ T_{X_ 2}(\omega_ s)\underline d^ T_{X_ 1}(- \omega_ s/b)\exp (-i\lambda a) \] where \(\underline d^ T_ s(\omega)\) is the Fourier transform of the signal S(t) at frequency \(\omega\), \(\psi\) is a weighting (empirical) function, \(\omega_ s=2\pi s/T\), \(s=0,\pm 1,... \). Under regularity conditions, it is proved that the derived estimate is consistent and asymptotically normal. Estimates of the spectra are also obtained, and applications to seismology are discussed in detail.