Adaptive quadratically-constrained MVDR filtering

The objective of this work is the development of receiver designs for wireless CDMA communications with a special emphasis given to high-dimensional adaptive processing under small sample support. In this work, we propose a quadratically-constrained MVDR/MMSE filter in an effort to achieve superior filtering performance under limited sample support adaptations. The quadratically-constrained MVDR/MMSE filter estimator utilizes a regularized estimate of the input covariance matrix that is given as a convex combination of the sample-average covariance matrix and a scalar multiple of its inverse. Selection methods based on the limited sample support are proposed for both the inverse matrix multiplicative scalar as well as the regularization parameter, and they can be applied more generally to other adaptive filtering techniques such as selection of the diagonally-loaded MVDR parameter.