Improvement of reconstruction accuracy for hardware systems with Finite Rate of Innovation

In classical Finite Rate of Innovation (FRI) theory, signal reconstruction is initially achieved with precision using the zeroing filter algorithm on noiseless pulse signals. However, with the introduction of noise, the zeroing filter algorithm exhibits reduced robustness when handling signals with low signal-to-noise ratios (SNR). This paper introduces the TLS-ESPRIT algorithm, which enhances robustness against noise to some extent. Additionally, building on the development of an FRI reconstruction system based on C-TEM sampling, this work innovatively proposes a cascaded filter phase compensation method. This method mitigates the nonlinear phase effects of analog filters, thereby improving reconstruction accuracy. Ultimately achieving a reduction in sampling rate while improving reconstruction accuracy.