State-space models for multichannel detection

In multichannel identification and detection (or model-based multichannel detection) problems the parameters of a model are identified from the observed channel process and the identified model is used to facilitate the detection of a signal in the observe process. A model-based multichannel detection algorithm was developed in the context of an innovations-based detection algorithm (IBDA) formulation for surveillance radar system applications. The state space model class was adopted to model the vector channel process because it is more general than the time series model class used in most analyses to date. An IBDA methodology was developed based on the canonical correlations algorithm which for state-space model identification offers performance advantages over alternative techniques. A computer simulation was developed to validate the methodology and the algorithm, and to carry out performance assessments. Simulation results indicate that the algorithm is capable of discriminating between the null hypothesis (clutter plus noise) and the alternative hypothesis (signal plus clutter plus noise). In summary, the applicability of the approach to radar system problems was established.