jpc@avdms8.msfc.nasa.gov (J. Porter Clark) (05/06/91)
(Something is eating about 50% of my posts. I *think* this one never made it out--please forgive me if you've seen it before.) I received a few responses to my "What is RELP" query. RELP stands for residually excited (or encoded, maybe) linear predictive code. The strain I've been hearing about digitizes voice at 9600 bps. I would have responded sooner but I lost one of the replies and spent some time looking for it. My need to know is this: The folks who are responsible for the space-to-ground link for Space Station Freedom are currently planning to use RELP to get two full duplex voice links at 9600 bps each per direction. I work on the SSF internal audio distribution and need to know how this affects our design. So, does anybody have any RELP code for sale or to be given away? +++++ Summary of responses +++++ The most complete discussion received to date and that I haven't managed to lose is from John Adcock at Lockheed Sanders. >From: adcock@rocket.sanders.com (john adcock) > >Yes that's what it stands for. > >Normal lpc vocoding uses only the all pole model of the >spectrum and the pitch to reconstruct the speech. That is to >say that during synthesis in a vanilla lpc vocoder the excitation >to the all pole filter is synthesized. RELP preserves the excitation >by saving the residual. That is with RELP you estimate >the pole model, then you do a zero filter with your pole model and preserve >the "residual" that you are left with. The residual is then used >to reconstruct the speech later. > >I've heard more about CELP >(Code excited linear predictive coding) which is >(as far as I know) based on the same idea but uses a >vector quantization method to code the residual instead of >actually transmitting the residual itself. Whether or not >a RELP or CELP vocoder is 9600 bps depends entirely on the >implementation. >NSA has a standard for the CELP algorithm and a 9600 baud >implementation. +++++++++++ >From: Rich Kerr <RJK@MILAN.QUALCOMM.COM> > >Try ICCASP 84,85. Presenter is PHil Wilson, Linkabit. > >Also PHD Thesis, Imperial University, London, Phil Wilson, 1982. RELP > ++++++++++ I also received a message from Tony Robinson (ajr@engineering.cambridge.ac.uk) referring me to books by Rabiner and Schafer and Jayant and Noll, but I had already gotten all I could get on the subject from those books. Thanks anyway. ++++++++++ Many thanks to all who responded. -- J. Porter Clark jpc@avdms8.msfc.nasa.gov