Pathwise MIMO channel modeling and estimation

A specular approach to time-varying frequency-selective MIMO channel modeling is proposed, that yields a parsimonious channel representation. Essentially, the vector of channel impulse response coefficients lives in a subspace that is subject to slow fading and hence is time-invariant over a reasonable time interval. The subspace is parameterized by the (deterministic) slow path parameters, that may comprise path delay, angle of departure and angle of arrival. The channel expansion into this propagation signature subspace yields coefficients that are subject to fast fading. They are modeled as independent lowpass autoregressive processes. We present an analysis of the identifiability of specular channels, and an algorithm achieving identification, as well as a performance evaluation over actual channel measurements.