Abstract:
In this paper, we model the noise present in middle and upper layers of the atmosphere for the data collected from the Indian MST Radar. People carry out their analysis assuming that the noise is Gaussian and in fact, in most of the scenarios, the noise is Gaussian. There is a much chance of getting inaccurate results if it is not. Gaussianity tests namely Autocorrelation (AC) and Power Spectral Density (PSD) tests are conducted to find whether the noise is Gaussian or not. In non-Gaussian cases, further analysis is carried out using Empirical Mode Decomposition (EMD). Once the exact type of noise contained in the data is known, specific denoising techniques can be applied so as to get better results. We develop the energy models of various noise distributions using EMD, test on random sequences, exponentials and derive the characteristics under various environments. Finally, the developed models are compared with the models obtained with the radar data and noise characterization is done.

Keywords: MST Radar, Noise Modelling, Wavelet Based Denoising, Principal Component Analysis, Gaussianity tests, Empirical Mode Decomposition, Intrinsic Mode Function