This paper presents a novel method for estimation of the joint probability
of multisensory signals by introducing dimension-reduction mapping function
s based on the principle of maximum entropy. A maximum mutual information c
riterion is derived for selecting the desired mapping functions. An algorit
hm is further presented for linear transformations of Gaussian random vecto
rs. Experimental results are shown to demonstrate the performance of the pr
oposed method. (C) 2001 Elsevier Science B.V. All rights reserved.