derest

Automatic smoothing and differentiation of a noisy time series

This is a code for off-line smoothing and estimation of first and second derivatives of a function from a sequence of noisy measurements, which can be nonequally spaced. The algorithm is automatic: the user does not need to provide smoothing parameters, they are estimated in the code. The signal is modelled as a stationary double-integrated Wiener process and estimates are computed using a Kalman smoother. Theoretical details of the algorithm are presented in https://arxiv.org/abs/1610.04397