Blind estimation of FIR channels using spatial separation

Abstract

Of the different information sources used in blind algorithms, explicit use of Finite Alphabet (FA) data is more recent than the use of either statistical data embedded at the source or algebraic information present in the channel itself. In channels that can be modeled using Finite Impulse Response (FIR) structures, the FA property results in the received vector set being a??clustereda?? around theoretical centers. These centers are the result of the convolution of the channel matrix with a given transmitter symbol constellation. This structure, together with the transition information of the received vectors, provides sufficient information for both symbol sequence and channel coefficient determination.The information generated as described above in a FIR Multiple Output channel results in the received data vectors forming clusters in a lattice like structure. Such, we refer to these data structures as spatial data. In this thesis, we present a family of algorithms that utilizes this spatial data for estimation. The algorithms presented include one direct symbol estimation algorithm and two channel estimation schemes. The two channel estimation algorithms process different aspects of the spatial data and thus exhibit a marked difference in their behaviors.