Reliable amplitude and frequency estimation for biased and noisy signals

Robust estimation of the amplitude, frequency and bias of unknown noisy sinusoidal signals is considered in this paper. It is only assumed that the measurements noise is bounded without any additional information such as stationarity, uncorrelation or type of distribution. In this context, the aim is to compute the set of all admissible values that are consistent with the measurements and with the error bound. The estimation problem is formulated as a Constraint Satisfaction Problem (CSP) where the amplitude, frequency and bias constitute the variables and a function relating them to the output is the constraint. Interval constraint propagation techniques are used to solve, in a guaranteed way, this problem. In order to illustrate the principle and the efficiency of the approach, numerical simulations are provided.

Research highlights
► Robust estimation of sinusoidal signals is considered.
► The noise is assumed bounded with know bounds.
► The set of all admissible values is computed.
► The estimation problem is formulated as a Constraint Satisfaction Problem.
► Interval constraint propagation techniques are used to solve this problem.