ABSTRACT
I. INTRODUCTION Blind deconvolution and equalization of linear time-invariant systems have widely received attention in various fields such as data communication, image processing, and geophysical data processing [1-5]. Recently, Shalvi and Weinstein proposed certain new criteria for blind deconvolution of nonminimum phase linear time-invariant systems [3]. Their work is, however, restricted to the single-channel (or scalar) case. Multichannel blind deconvolution is closely related to multichannel blind signal separation [5-9]. Multichannel blind deconvolution usually assumes that the components of input signals are white in the second-or higher-order sense [4], but multichannel blind separation lacks such an assumption as whiteness, and assumes that the components of input signals are mutually independent [5-9]. Most of the approaches proposed for blind signal separation are developed under the assumption that the linear systems involved are unknown constant gains [5-8]. Yellin and Weinstein proposed new criteria for multichannel blind signal separation [9]. However, their approach is confined to a restrictive case in which all the diagonal elements of the transfer function matrices involved are constant and equal to unity.
