This article deals with the issue of optimum detection of a known spread spectrum (SS) signal in nonwhite noise and/or narrowband interference. Prior to detection, the input signal is whitened according to the reciprocal of an interfering spectrum, I(z)=1/AG(z). Next, in force of the maximum likelihood (ML) principle, the replica of signal is carried through the same filter, I(z), to match it to a new signal shape. Results of simulation show a significant gain in comparison to conventional single-matched detection.