Gravitational Waves have been looked for over the last forty years without success. Predicted by the General Relativity, they are analogous to Electromagnetic Waves, except the fact that there are no negative masses and therefore the radiation is reduced to its quadrupolar emission, not to speak of the very weak gravitational forces compared to the electric ones! Traditional analysis of Gravitational Waves (emitted by Black Holes or Neutron Stars) is made through Fourier transforms and its improved variants. These algorithms work fine provided the gravitational signal is not totally drawn inside the noise, which is unfortunately the case. So, one has to find a very tiny needle (the signal) inside a huge haystack (the noise). We will show that this is feasible using the fact that the signal and the noise have totally different analytic properties in the complex plane. Another way to present the point is to say that the noise is not uniformly distributed in the Complex Plane. There are noise attractors with roots of unity and staying away from these attractors allows the capture of the signal.