Blind identification and equalization for multiple-input/multiple-output channels

Abstract

To separate and recover multiple signals in data communications, antenna arrays and acoustic sensor arrays, the impulse responses of multiple-input/multiple-output (MIMO) channels have to be identified explicitly or implicitly. This paper deals with the blind identification of MIMO FIR channels based on second-order statistics of the channel outputs and its application in blind channel equalization and signal-separation. We first investigate the identifiability of MIMO FIR channels, and obtain a necessary and sufficient condition for MIMO FIR channels to be identifiable up to a unitary ambiguity matrix using second-order statistics. Next, we extend the identification algorithms for single-input/multiple-output (SIMO) FIR channels, such as the algebraic algorithm and the subspace algorithm to the identification of MIMO FIR channels. Since the ambiguity matrix cannot be estimated using second-order statistics, we then present a fourth-order cumulant based ambiguity matrix estimation algorithm. Finally, we demonstrate effectiveness of the algorithms in this paper by computer simulations.