🚀 go-pugleaf

christos@kuhub.cc.ukans.edu () writes:

> Hello there,

> I am using the Public Domain program Sound Tools on the Sun station
> to record speech. The speech samples are saved in a mu-law 8bits/sample
> encoding rule with a standard ".au" header precceding the data. What I
> would like to know is what the format of the speech samples is? In other
> words, I am reading the bytes that correspond to 8-bit speech samples but I
> don't know whether they are in signed binary format(2's complement) or
> unsigned binary format or anyother format. I have tried both binary signed
> (2' complement) and unsigned but it doesn't work. Does anybody know how the
> samples are stored?

> I would appreciate any relevant response.

The most likely encoding will be a standard telecomms encoding at 64kbps or
8kHz audio. One second of information will be 8000 bytes, not 8kbytes.

The 'original' samples are 12 bits wide and a compression algorithm is
applied to get down to 8 bits. The USA (and dependants) use the mu-law
encoding scheme. We Europeans use an A-law scheme.

So much for background. If I remember correctly, the most significant bit is
a sign bit, indicating positive (0) or negative (1) value. The remaining 7
bits are the encoded amplitude of the sample. Manipulation will require you
to decode the sample correctly, and I don't have the details of mu-law
encoding to hand. But, I know where it is described. Try and get hold of the
CCITT's G.7xx series recommendations which is one large volume. Try a
technical library of a university. Alternatively, you could ask very nicely
and I will try to mail the relevant pages when I'm over in the US in January.

Richard Spence
rspence@cix.compulink.co.uk
CompuServe: 100112,304

>>>MATRIX version  1.21e