Combination Of Time-Frequency Representations For Background Noise Suppression

The aim of the paper is to propose approach for enhancement of time-frequency representation leading to the background noise suppression. The approach is based on combination of continuous wavelet analysis, timevarying autoregressive process and short time Fourier transform. By such combination we make the identification of important areas in the time-frequency representation easier. The proposed method is an alternative approach to significance tests which can be problematic in some cases. The performance of methods is presented on the gross domestic product of the United Kingdom and Group of 7. The results show that in the UK, oil crisis has a bigger impact compared to financial crisis, while from the perspectives of G7 countries, the impact of financial crisis was stronger. The obtained results can be also used for consequent econometric analysis which identify dependencies, relations, bilateral causalities or other economic aspects.