| Wireless communication systems are witnessing a rapid growth in the number of subscribers and in the range of services. An important challenge for wireless systems today is the limited availability of the radio frequency spectrum. Antenna arrays offer a promising approach for increasing the spectrum efficiency of multiple access wireless networks. The spatial discrimination provided by antenna arrays allows multiple users located in different directions to share the same channel. In this thesis, we focus on the signal processing foundations of this emerging technology. Specifically, we develop blind space-time algorithms for estimating multiple co-channel user signals received at an antenna array.;We assume that the user signals are digitally modulated with the same symbol rate and alphabet. The signals arrive at the array through a multipath propagation channel. The array output is a superposition of signals from all the users, and additive noise. Our objective in blind space-time signal processing is to estimate the channel and information bits of each user, without the aid of training sets. We achieve this goal by exploiting the finite alphabet (FA) property of digital symbols.;The blind estimation problem is first studied for memoryless channels with negligible delay spread. The data model yields a structured matrix factorization problem of the form X = AS, where X is a matrix of array outputs, A represents the unknown spatial channel, and S represents the unknown FA symbols. We study the uniqueness of this factorization, propose computationally efficient algorithms for estimating the factors A and S, and analyze the convergence and error performance of the algorithms.;We next consider multipath channels with large delay spread. In this case, each antenna output is temporally oversampled at a rate higher than the symbol rate. This results in a similar, but larger matrix factorization problem X = HS, where H represents the unknown space-time channel, and the symbol matrix S now has a block-Toeplitz structure. We show how to reduce this problem to the previous one by exploiting the Toeplitz structure of S. The effectiveness of this approach is confirmed by simulation studies. |
