breck@ganzer.ecs.umass.edu (Liam Breck) (04/27/91)
A little while back I posted: > Can someone please sugest references to DETAILED descriptions > of how to implement ADPCM sound compression. Here is a summary of the replies... Suggestions: One "good" way of coding sounds, lossless, to get decent compression is to change the absolute representation to a delta ditto. Even if the absolute location of a sound wave isn't the same from time to time, the shape is, and it's the shape that's represented by the differences. The differences could then be compressed using either a LZ-like approach or simply coded by Huffman or Arithmetic Coding. If you're into that kind of thing I presume a DCT (Discrete Cosine Transform) in one dimention would give you an OK result, since that's the kind of stuff that's at the bottom of a sound anyway :-) One thing I've wanted to do, but haven't done yet, is a kind of curve fitting; a slope factor, a curve factor and a range factor maybe. This could be improved to allow for literal chunks. --- If you start with 8 bit numbers then this is not very high quaility anyway, so a simple difference operator followed by Haufmann coding would probably give you 2 or more times compression. References: Digital Coding of Waveforms by N.S. Jayant and P. Noll Prentice Hall Inc. Englewood Cliffs, NJ 1984. ISBN 0-13-211913-7 01 ...covers the basics in signal compression. Jayant was one of the main contributors to the theory of ADPCM. --- Bellcore TR-TSY-000499, Transport Systems Generic Requirements: Common Requirements. ...a very detailed description of the CCITT standard ADPCM algorithm Call Bellcore Customer Service to get a copy: 201-699-5800 Cost $15 - 30 ? Their catalog is free, and it lists zillions of technical reports, as well as things like test tapes of voice, disks with uncompressed and compressed voice to test your compression algorithm, etc. Prices range from cheap to very expensive, especially for things with information on disk, or multi-thousand-page documents. --- "Digital Processing of Speech Signals" by Rabiner & Schafer Prentice-Hall 1978 ISBN 0-13-213603-1 ...covers the gamut from simple delta modulation to various LPC techniques. Pretty math-intensive, but also include some good empirical data on the effect of various distortions as caused by the various encoding schemes. --- Speech Communications by O'Shawnessy (O'Shaughnessy?) circa 1989 -- Liam Breck breck@umvlsi.ecs.umass.edu