A generalized framework for development of partially-updated signal and parameter estimation algorithms based on subspace optimization constraints

A generalized framework for development of subspace constrained partially-updated (SCPU) signal and parameter estimation algorithms is proposed and demonstrated via analysis and computer simulation. Conventional partial-update (PU) methods are first reviewed and interpreted as a sequence of cost function optimizations subject to a hard parameter constraint. The SCPU method is then introduced as an equivalent optimization subject to a soft subspace constraint. It is shown that the new method removes adaptive misadjustment inherent to conventional PU methods, and allows generalization of the partial-update methods to much broader classes of signal and parameter estimation algorithms, including blind and nonblind ML estimation methods.