bhine@pioneer.arc.nasa.gov.arpa (Butler Hine sst) (12/09/88)
[gobble] I kept the sample of NeXT voice mail sent out a while ago, but I never saw any explanation of the format. Does anyone know what the format is? I saw a note from someone saying they had decoded it, but I can't find their address. Help! Thanks in advance. Butler Hine NASA Ames Research Center hine@galileo.arc.nasa.gov
desnoyer@Apple.COM (Peter Desnoyers) (12/09/88)
In article <19355@ames.arc.nasa.gov> bhine@nike.UUCP (Butler Hine sst) writes: >[gobble] > >I kept the sample of NeXT voice mail sent out a while ago, but I never >saw any explanation of the format. Does anyone know what the format is? >I saw a note from someone saying they had decoded it, but I can't find >their address. Help! Thanks in advance. > > Butler Hine > NASA Ames Research Center > hine@galileo.arc.nasa.gov I posted that message. The format is simple, like uuencoded data but just a hair different. The binary data is standard mu-law PCM voice at 8k samples/second, with 8 bit samples. the ascii format is lines of: <space>abcdabcd...(64chars) where each group of 4 characters ('abcd') decodes to 3 bytes as follows: subtract 33 (decimal) from each character to get a 6-bit value. Concatenate 4 6-bit values to get 3 8-bit values. Peter Desnoyers
rminnich@super.ORG (Ronald G Minnich) (12/13/88)
In article <21949@apple.Apple.COM> desnoyer@Apple.COM (Peter Desnoyers) writes: >just a hair different. The binary data is standard mu-law PCM voice at ^^^^^^^^^^ Is it possible to explain this in, say, less than 500 words. An equation maybe? This is outside of my domain ... Thanks, ron
wcs@skep2.ATT.COM (Bill.Stewart.[ho95c]) (12/14/88)
In article <2062@super.ORG> rminnich@duper.UUCP (Ronald G Minnich) writes: :In article <21949@apple.Apple.COM> desnoyer@Apple.COM (Peter Desnoyers) writes: :>just a hair different. The binary data is standard mu-law PCM voice at : ^^^^^^^^^^ :Is it possible to explain this in, say, less than 500 words. Well, to start with, PCM is Pulse Code Modulation - you sample the analog waveform at N samples per second (telephone business uses 8000 samples/sec) and send a digital code representing the amplitude of the signal. There are two standard encodings around - USA & Japan use mu-law and Europe uses A-law. Both have basically the same approach - 8 bits of data, with a non-linear representation. Because human hearing is non-linear, you preserve the most sound fidelity if you represent low amplitudes more precisely, and higher ones less precisely. So the change in sound level between byte values 2 and 3 is much smaller than the change between 126 and 127. I don't have my formulas handy, but the basic difference between mu-law and A-law is whether there's a code to represent 0, or whether the encoding is symmetric, with +/- epsilon represented by codes 0 and -1. -- # Thanks; # Bill Stewart, AT&T Bell Labs 2G218 Holmdel NJ 201-949-0705 ho95c.att.com!wcs # # News. Don't ask me about News.
desnoyer@Apple.COM (Peter Desnoyers) (12/14/88)
In article <2062@super.ORG> rminnich@duper.UUCP (Ronald G Minnich) writes: >In article <21949@apple.Apple.COM> desnoyer@Apple.COM (Peter Desnoyers) writes: >>just a hair different. The binary data is standard mu-law PCM voice at > ^^^^^^^^^^ >Is it possible to explain this in, say, less than 500 words. >An equation maybe? This is outside of my domain ... >Thanks, >ron Sure. Mu-law encoding is what everyone uses (e.g. Sprint) unless they're really cheap. (Note that almost all telephone transmission except the local loop to your phone is now done digitally, usually in this format.) The way it works can be viewed thus: Sample voice at 13 bits per sample, 8000 times per second. Apply a sort-of-logarithmic function to squeeze these 13 bits into 8: +1 -> +1, while +4096 -> +128. Unsqueeze it at the destination. Thus your sampling noise is very small for quiet signals at the cost of increased sampling noise for loud signals. The trade-off must be optimal - Bell Labs spent years researching it. (1/2 :-) The function is actually a piece-wise linear approximation to something sort of logarhythmic, most likely because it was easier to do back in the 60's when they first started using digital transmission. (Yup, even then it was cheaper to spend _lots_ of money on digital hardware than it was to find space to string more wires underground in Manhattan. I think it was the first application of transistors in the public network.) [I apologize for any technical inaccuracies or vagueness. I couldn't find any of my references on mu-law. I have a mu-law to linear table on line, but the machine is down. oh well...] Peter Desnoyers
keith@prism.gatech.EDU (Keith Edwards) (11/21/89)
Forgive me if this has already gone by before, but is there any documentation available (from NeXT or otherwise) which describes the VoiceMail format? I've dug through the online documentation but can't find anything. Thanks, Keith -- keith edwards -- the software engineering research center georgia tech internet: keith@gatech.edu atlanta, ga uucp: {the_known_world}!gatech!keith 30332-0280