Learning strategies for adaptive equalizers using the constant modulus error criterion

In communication systems, data are often corrupted during transmission due to imperfect communication channels. Since channel characteristics may be time varying or unknown prior to data transmission, adaptive equalizers are typically incorporated into the receiver to reduce the ill effects of the imperfect channel. Conventional equalizers utilize a training signal to achieve the proper correction. Blind equalization schemes do not require the use of a training signal, but instead attempt restoration based upon some known property of the transmitted signal. The constant modulus algorithm (CMA) is a blind equalization technique that may be used to equalize certain communication signals (e.g. BPSK, QPSK, and FM). This paper investigates three new techniques for adaptive equalization based on the constant modulus error criterion: (i) the fast quasi-Newton CMA, (ii) the conjugate gradient quasi-Newton CMA, (iii) the conjugate gradient CMA. These three methods are formulated analytically and evaluated experimentally on several typical communication channel models.