A spherical subspace tracker with adaptive rank estimation for detection of phase-modulated signals

Abstract

In this thesis, detection algorithms for cellular wireless communication are presented, which track the dominant eigenvalues and eigenvectors of the multipath fading channel. A decision variable may be formed from these eigenvectors and eigenvalues which uses the multipath diversity inherent in the channel, as well as advantageously combines orthogonal (uncorrelated) signal components. The determination of the rank of the signal-noise subspace is effected in the algorithms in two different ways; in the first, by calculation of an information criterion giving the probable dimension, and in the second, by choosing the dimension which includes a given percentage of the signal energy. Realistic simulations using measured channel impulse response data show a small performance gain of the eigenanalysis detection algorithm over simple differential detection.