This paper presents a new iterative equalization algorithm that maximizes c
apacity for discrete multitone (DMT) systems. The research modifies a previ
ously proposed criterion and applies an appropriate transformation to map t
he objective function and the constraint set into a canonical region. The r
esulting constraint set exhibits an identifiable geometric characteristic.
Using the gradient projection method in conjunction with projection onto co
nvex sets (POCS) provides us with an iterative search algorithm that facili
tates the gradient descent method. We also generalize the approach to two i
mportant subclasses of equalizers, namely linear phase and unit tap filters
. We also derive a fundamental limit on the performance of the proposed app
roach. In comparison with the previous methods, the proposed equalization a
lgorithm is less computationally complex and more geometrically intuitive.
Simulation experiments confirm the validity of the proposed method for equa
lization of DMT systems.