In this paper the author considers a general method, based on time domain samples for spectral manipulation of time limited signals. First, the original signal is divided into some frames in the time domain. Then, by presenting a suitable theoretical and computational algorithm, and using a method for improving the speed of convergence, we find the local bandwidth of each frame; thereby, each frame is sampled with a new rate, so the amount of data per frame is decreased. Finally, by storing suitable information at the beginning of each frame, the original signal could be reconstructed. Therefore, by this method some redundant data is removed and the net volume of data is reduced. The reconstruction of original signal is achieved by using periodic interpolation. In this technique, a method for reduction of error (due to false interpolation) at the end of any frame is examined carefully.