Multi-resolution dictionary learning for discrimination of hidden features: A case study of atmospheric gravity waves

In this work, we propose Multi-Resolution Dictionary Learning by promoting sparsity, and Instantaneous Frequency Estimation technique for the discrimination of measured parameter from the effect of unknown sources like noise, and natural phenomena that occur in background during the data acquisition. The Dictionary Learning technique using Multi-Resolution Analysis is performed in the Wavelet Analysis domain using Singular Value Decomposition. In this paper, we aim for the discrimination of contribution from unknown sources in the measured parameter in the amplitude domain and frequency domain by promoting sparsity and Instantaneous Frequency Estimation respectively. We implement the proposed technique and discuss its potential with a case study on the discriminating the Doppler shift in Atmospheric Gravity Waves that occur due to the horizontal wind during data acquisition.