ralphw@ius3.ius.cs.cmu.edu (Ralph Hyre) (05/23/88)
[This is a slightly edited copy of my rec.arts.int-fiction post, since this discussion thread started there.] Does Anyone know of an analog<->LPC<->Audio chip (or chip set) where the input LPC parameters are the same as the output paramaters? (I don't want to have to build a custom DSP system just to have what amounts to a vocoder.) In article <4556@dasys1.UUCP> newsome@dasys1.UUCP (Richard Newsome) writes: > >The latest INFOWORLD has an article on adding sound to Hypercard stacks .... >Does anyone know anything about the possibilities of sound >compression? My text compressor seems to store an eight-bit ASCII >character in 4 bits; and according to Claude Shannon the theoretical >limit for English text is about 1 bit per character For speech, people have had good results with LPC (Linear Predictive Coding) [I believe it's a measure of energy at particular wavelengths] The TMS 5220 chip sounds OK with 1200 bps data, but last I checked, generating the LPC paramaters for it basically required a $10,000 development system from TI. General Instruments makes a LPC chip (SP-01 I believe) which take converts back and forth between LPC and analog signal, but I recall reading in Byte [in a Steve Ciarcia construction article for the Lis'ner 1000 (speech recognition system)] that the input parameters aren't the same as the output paramaters. So it seems the chip makers are trying to generate some extra revenue at the expense of hardware hackerbeing able to do real things with these chips. BBN developed some 300 bps technology, but it's optimized for voice, so music won't sound right. Sorry that I don't have more concrete references. -- - Ralph W. Hyre, Jr. Internet: ralphw@ius2.cs.cmu.edu Phone:(412)268-{2847,3275} CMU-{BUGS,DARK} Amateur Packet Radio: N3FGW@W2XO, or c/o W3VC, CMU Radio Club, Pittsburgh, PA