Effect of signal bandwidth on the accuracy of adaptive interpolation of discrete-time signals

In this paper, we study the effect of signal bandwidth on the interpolation of discrete time signals that could be modeled as autoregressive processes. Recently, interpolation of discrete-time signals has received attention from researchers and a variety of interpolating techniques have been used with varying degrees of success. If the signal could be modeled as an autoregressive process, then such a model could be used in signal restoration. The scheme has to be adaptive in nature as in any practical situation, the model parameters have to be estimated from available data. First, we obtain some general results regarding the interpolation error as a function of bandwidth for first order and second order AR processes. It is shown that the interpolation error is a function of the bandwidth of the process, and in general, narrowband processes give lesser interpolation error than broadband processes. Then we use simulation to verify the theoretical results obtained. The simulations are carried out for cases when the misses occur in random as well as when the misses occur in bursts.