Sequential Detection Estimation and Noise Cance- Lation

A noise canceling technique based on a model-based recursive processor is presented. Beginning with a canceling scheme using a reference source, it is shown how to obtain an estimate of the noise, which can then be incorporated in a recursive noise canceler. Once this is done, recursive detection and estimation schemes are developed. The approach is to model the nonstationary noise as an autoregressive process, which can then easily be placed into a state-space canonical form. This results in a Kalman-type recursive processor where the measurement is the noise and the reference is the source term. Once the noise model is in state-space form, it is combined with detection and estimation problems in a self-consistent structure. It is then shown how parameters of interest, such as a signal bearing or range, can be enhanced by including (augmenting) these parameters into the state vector, thereby jointly estimating them along with the recursive updating of the noise model.