A reduced sufficient statistics-based algorithm for joint timing/channel estimation in blind equalization

In blind equalization, both the symbol timing and channel coefficients are unknown a priori. Previous algorithms for joint estimation of these parameters have used the extended Kalman filter, which is subject to divergence at low SNRs. Here, we present a new joint estimation algorithm which is based on the reduced sufficient statistics (RSS) method of Kulhavy (1990). The resulting channel and timing estimator is shown to use a modified recursive least-squares algorithm for the channel coefficients, and a joint nonlinear multiple-model type estimation for the timing. The application of the RSS estimator to blind symbol-by-symbol detection (SBSD) is illustrated.