Programmable canonical correlation analyzers with recursion and feedback

Modified programmable canonical correlation analyzers (PCCA) are developed to exploit recursion and feedback for improved blind adaptive spatial filtering. Specific implementations are developed utilizing an alternating block power method with a generalized Gram-Schmidt orthogonalization procedure. Several realization of these new recursive/feedback PCCAs are developed for exploitation of cyclostationarity and constant modulus signal properties. The performance of the proposed techniques is evaluated empirically and characterized in terms of output SINR and convergence behavior.