Multipath parameter estimation of time-varying space-time communication channels using parallel factor analysis

Abstract

In this work we propose a new method for estimation of multipath parameters in the context of mobile communications. The proposed method is an effective way to exploit the fact that the paths amplitudes are fast-varying while angles and delays are slowly-varying over multiple time-slots. By relying on such a multislot invariance of angles and delays and by periodically extending a training sequence over multiple time-slots, we show that the received signal can be modeled as a third-order (3D) tensor, which follows a Parallel Factor (PARAFAC) model. An accelerated alternating least squares (ALS) algorithm is used for a joint estimation of the angles of arrival, delays and fading amplitudes of the multipaths. Numerical results from computer simulations show that the proposed PARAFAC-based estimator is capable of estimating the multipath channel parameters with good accuracy using short training sequences and with fewer receiver antennas than multipaths.