Blind equalization and identification of communication channels

In this dissertation, we consider direct channel identification and blind equalization of both SIMO and MIMO channels using only (or primarily) the second-order statistics (SOS) of the data. Such channels arise when a single receiver data from a single user or multiple users is fractionally sampled or when an antenna array is used with or without sampling or when both these scenarios are applicable.; Two major sources of impairments of single-user digital communications signals as they propagate through channels are multipaths and limited channel bandwidth. This leads to intersymbol interference (ISI) at the receiver which, in turn, may lead to high error rates in symbol detection. In multi-user (multi-access) communications the signals from the other users act as interferences. Equalizers are designed to compensate for these channel distortions and to reject multiaccess interferences. One may directly design an equalizer given the received signal, or one may first estimate the channel impulse response and then design an equalizer based on the estimated channel. Typically a training sequence (known to the receiver) may be transmitted during start-up (acquisition mode). In the operational stage, the receivers typically switch to a decision-directed mode where the previously equalized and detected symbols are used as a (pseudo-)training sequence together with the received data to update the channel or the equalizer coefficients. Recently there has been much interest in blind (self-recovering) channel estimation and blind equalization where no training sequences are available or used. In multiaccess systems the training sequences must also be provided by the interference-generating sources, an utterly unrealistic assumption.; The mathematical model for multiple user communication systems turns out to be a multiple-input multiple-output (MIMO) discrete-time linear system. For single-user systems the underlying mathematical model is a single-input multiple-output (SIMO) discrete-time linear system. Therefore, in this dissertation we address the problem of blind channel estimation and blind equalization of both SIMO and MIMO channels using primarily the SOS of the data. Estimation of (partial) channel impulse response and design of finite-length MMSE (minimum mean-square error) blind equalizers is investigated. Four approaches (two based on a whitening concept and two based on linear prediction) are considered. Unlike past work, we allow a much broader class of channel models. For instance, common “subchannel” zeros are allowed, infinite impulse channels are allowed, channel impulse response length or model order need not be known, and in the case of MIMO systems, the MIMO channel transfer function need not be column reduced. All these restrictions in the existing approaches to this problem have precluded wider applicability of blind techniques to digital communications channel equalization.; Illustrative computer simulation examples are presented in support of the proposed approaches.