Although the reduced-state technique has been widely used in Viterbi equali
sers, its application to maximum a posteriori probability (MAP) equalisers
is not satisfactory because it cannot operate due to the backward recursion
of the conventional MAP algorithm. A new kind of MAP algorithm is proposed
in which the forward and backward recursions are exactly symmetric. The re
duced-state technique can thus be applied to both forward and backward recu
rsions and a gain in performance is achieved over that of conventional redu
ced-state MAP equalisers at the cost of double the computational overhead.