Abstract:

Atmospheric noise must be suppressed in low-frequency communication systems. This paper studies the amplitude characteristics of atmospheric noise based on real data and proposes a nonlinearity method. Firstly, Analysis on real noise amplitude demonstrates the non-Gaussianity of atmospheric noise and shows that the amplitude roughly follows the α-stable distribution. Then, based on the traditional Gaussianization algorithm, this paper deduces a novel nonlinearity method which is called as the over-Gaussianization transformation. Finally, simulations of normality test and bit error ratio (BER) on real noise show that higher Gaussianity does not always mean lower BER. The over-Gaussianization method achieves similar BER performance to the locally optimal detector.