“…Such extensions have been proposed for FS-1016, MELP, G.723.1 [5], G.711 [6], G.729 [8], and AMR-WB [8], [7], [9]. The common thread throughout this work is the insightful division of the standardized (or slightly modified) encoder output into m descriptions so that when taken together, those descriptions produce speech with a quality near that of the standardized coder, and when fewer than m descriptions are available, they can still be decoded to produce somewhat lower-quality speech.…”