The S transform is a type of time-frequency analysis transform composed of some concepts of short time Fourier transforms (STFT) and wavelet transform. To localize time-frequency spectrum in this method, a scalable Gaussian window is used. The scalable Gaussian window provides time-frequency resolution while maintaining a direct relationship with the Fourier spectrum. In this study, after introducing the S transform, generalized S transform is introduced and its application in first arrival time estimation of noisy seismic records is presented.