The stability of a procedure for the recovery of lost samples in band- limited signals (Q1341382)

The paper deals with the problem of stability of a procedure for the recovery of lost samples in band-limited signals. Theoretically, for band-limited signals, the positive-definite character of reconstruction matrix \(S\) guarantees the existence and uniqueness of the solution of the reconstruction problem. But in practice, determining vector of unknown samples, a number of difficulties may occur. In the paper the eigenstructure of matrix \(S\) is clarified. Theorems are proved about the bounds of eigenvalues of \(S\) as a function of oversampling parameter \(r\). The bounds, given in the theorems are independent of the number of missing samples \(n\), and are extremely simple to calculate. It is proved that the largest eigenvalue of \(S\) is always larger than \(r\), and \(S\) may reach values nearly as low as \(r\), independently of the number of lost samples. Several useful bounds to the eigenvalues of \(S\), and examples which illustrate and confirm the main results, are given. For solving the reconstruction problem, it is shown that it is possible to formulate a reconstruction problem similar to the one studied for finite-dimensional signals. The presented results may have practical applications, some of which are apparent. In general, the paper is addressed to scientists working in the field of stochastic processes, especially stochastic processes reconstruction.