No, I dont get that read. You need a form of matched filter but that's software. You use a pseudo random code to encode the signal and at the other end a matched filter (can be hardware or just software depending of the speed/bandwidth needs) which is properly seeded with the same pseudo random code sequence can pluck out the correct signal componenets from where they were buried. This is done all the time with spread spectrum radios, radars, etc. I admit I havent read the paper but from the article description this sounds like basic spread spectrum technology applied to a fiber optic channel. Different noise characteristics than the atmosphere, which is probably why this is considered novel, but the technique is well established and in use all over.
Interestingly, Spread Spectrum was first patented by Hedy Lamarr (the actress!) in the early 1940s.