The picture seems to match what the article says, with respect to a detection system having a step-like threshold. It is not a matter of the signal being detected even in a high-noise environment, as in this case, it would not be detected at all without the noise: the signal (blue) line is always below the threshold (the dotted line.) With noise added, the detector picks up the spikes above the threshold (those circled in red), and the time-sequence of those spikes has the same periodicity and phase as the original signal (plus some noise).

The article points out that with too much noise, this does not work.