Reconstruction from Non-uniform Samples

Exact reconstruction of a band-limited signal from its non-uniform samples involves the use of Lagrange interpolation, which is impractical to implement as it is computationally difficult. This thesis develops approximate reconstruction methods based on time-warping to obtain reconstruction of band-limited signals from non-uniform samples. A review of non-uniform sampling theorems is presented followed by an alternative interpretation of the Lagrange interpolation kernel by decomposing the kernel into its constituent components. A discussion of time-warping and its use in the context of non-uniform sampling is made. This includes an alternative interpretation known as the delay-modulation, which we show to be a simpler representation for a specific case of non-uniform sampling where the sample instants are deviations from a uniform grid. Based on some essential characteristics of the Lagrange kernel, a framework using a modulated time-warped sine function is formed to obtain various approximations to the Lagrange kernel. The thesis also formulates a vector space representation of non-uniform sampling and interpolation and incorporates warped sinc functions to obtain faster convergence in iterative algorithms for reconstruction of band-limited signals from non-uniform samples. Thesis Supervisor: Alan V. Oppenheim Title: Ford Professor of Engineering