Conventional time and space filtering of stationary random signals, wh
ich amounts to forming linear combinations of time translates and spac
e translates, exploits the temporal and spatial coherence of the signa
ls. By including frequency translates as well, the spectral coherence
that is characteristic of cyclostationary signals can also be exploite
d. This paper develops some of the theoretical concepts underlying thi
s generalized type of filtering called FREquency-SHift (FRESH) filteri
ng, summarizes the theory of optimum FRESH filtering, which is a gener
alization of Wiener filtering called cyclic Wiener filtering, and illu
strates the theory with specific examples of separating temporally and
spectrally overlapping communications signals, including AM, BPSK, an
d QPSK. The structures and performances of optimum FRESH filters are p
resented, and adaptive adjustment of the weights in these structures i
s discussed. Also, specific results on the number of digital QAM signa
ls that can be separated, as a function of excess bandwidth, are obtai
ned.