Abstract: Digital audio or sound information such a speech is typically transmitted in bursts or packets containing a plurality of digitized sound samples. For the reconstruction of such a digital data stream in the presence of erasures or mutes, a method and an arrangement is disclosed, in which at a data receiver having an audio processing portion a plurality of successively data bursts outputted from a speech buffer are stored into a cyclic buffer. During an erasure, the data stream is reconstructed by outputting into the speech buffer stored bursts in their reversed order of storage. By reusing the bursts while reversing the sequence of the digitized sound samples, a virtually click free reconstruction of the data stream is obtained. An embodiment for use with a cordless telephone system is described.