Quadratic-linear detection of broadband non-Gaussian signals

The problem considered is that of detecting a possibly non-Gaussian signal imbedded in Gaussian or non-Gaussian noise. Broadband data is of principal interest. A well-known measure of performance is the deflection, or detectability index. The solution of the optimum detection operation under the deflection criterion is well-known for the case of a zero-mean non-Gaussian signal in Gaussian noise when a quadratic operation is used. This solution has been applied to the design and analysis of sonar arrays. Results for more general formulations of this problem are given, including the optimum operation when the data contains a non-zero mean and the operation is the sum of a quadratic and a linear functional, along with an analysis of a more general deflection-based criterion