Filterbank implementation of a window based Gabor transform

Abstract

Joint time-frequency analysis (JFTA) can be used in applications such as speech, music, radar, and sonar, where the frequency content of a signal during successive time intervals must be determined. Gabor (1946) presented a JFTA technique where a signal could be described as the sum of shifted and modulated Gaussian basis functions which are weighted by complex constants. Further developments have permitted the practical implementation of the Gabor transform on digital computers. However, these techniques were not well suited for real time processing of digital sequences of very long or of infinite length due to the requirement to receive and store the entire signal prior to processing. A solution to this is presented in the form of a window based Gabor transform (WBGT), which is suitable for filterbank implementations. It is developed based on the orthogonal like discrete Gabor transform (OLGT) and is implemented in filterbanks.