On short-time Fourier transform with single-sided exponential window

A special case of the short-time Fourier transform with single-sided exponential window, named exponentially forgetting transform (EFT), is discussed. It is shown that the EFT offers in a certain sense good time-frequency resolution and that stable reconstruction of a signal from samples of the EFT at equidistant time-frequency grid points (nT, k Omega) is possible, even for the case of nonredundant sampling (T Omega = 2 pi). Finally, computationally efficient algorithms for analysis and synthesis of discrete-time signals are developed. Computational efficiency is gained by using recursive computation in combination with the FFT algorithm.