This work considers the problem of joint blind channel equalization and co-channel user separation for wireless digital communications. We first show that some blind source separation techniques can be modified and applied to equalize channels and separate users in one step. The most important advantage of the general approach is the robustness to channel length mis-specification and channel ill-conditioning. We then propose the so-called fixed-window constant modulus algorithm with both batch and adaptive implementations. It overcomes the ill-convergence problem encountered by existing adaptive constant modulus algorithms, and thus greatly speeds up the convergence for use in a time-variant mobile environment. In the last part, we propose a blind channel identification approach for a single-user single-channel system where a cyclic prefix is inserted into the transmitted sequence. The proposed method, which is based on second-order statistics, allows the channel to be longer than the cyclic prefix and the block synchronization to be unknown.
展开▼
