Symbol-rate blind equalization based on constrained blind separation

Blind adaptive equalization has been of long-standing interest, and Bussgang-type algorithms for blind equalization based on gradient descent schemes are well-known. Algorithms based on blind source separation (BSS) have also been of interest. These exploit the independence of a transmitted symbol sequence, and are generally based on a fractional sampling scheme. In this paper, we study BSS based algorithms operating under symbol-rate sampling. With a Toeplitz structure constraint on the separating matrix, the source signal can be recovered with faster convergence. We also show that signal phase recovery is possible under an I/Q constraint, or with proper choice of a quantized nonlinearity.