Title:A Generalized Nyquist-Shannon Sampling Theorem Using the Koopman Operator

Abstract:The sampling theorem plays a fundamental role for the recovery of continuous-time signals from discrete-time samples in the field of signal processing. The sampling theorem of non-band-limited signals has evolved into one of the most challenging problems. In this work, a generalized sampling theorem -- which builds on the Koopman operator -- is proved for signals in generator-bounded space (Theorem 1). It naturally extends the Nyquist-Shannon sampling theorem that, 1) for band-limited signals, the lower bounds of sampling frequency given by these two theorems are exactly the same; 2) the Koopman operator-based sampling theorem can also provide finite bound of sampling frequency for certain types of non-band-limited signals, which can not be addressed by Nyquist-Shannon sampling theorem. These types of non-band-limited signals include but not limited to, for example, inverse Laplace transform with limited imaginary interval of integration, and linear combinations of complex exponential functions. Moreover, the Koopman operator-based reconstruction algorithm is provided with theoretical result of convergence. By this algorithm, the sampling theorem is effectively illustrated on several signals related to sine, exponential and polynomial signals.