Study of transforms and methods for their comparison

Information extraction from signals has been a long time research topic and numerous signal transforms have been defined for the same purpose. In this paper, a comparative study of various signal transforms on basis of time-frequency (TF) resolution, cross-terms suppression and maximum information content has been presented. The transforms considered for the analysis are Frequency domain transforms, Linear time-frequency transforms and Cohen's class distributions. All transforms have been compared visually w.r.t. to their ability to describe transformed plane. Additionally, two quantitative measures namely Renyi entropy and pattern recognition accuracy achieved while using individual transforms for getting features have also been used for the comparison. The whole study comprises of analysis of synthetic chirp signals and real time acoustic signals. The case studies' results have showed a direct correlation between Renyi Entropy and Pattern recognition accuracies.