[geeks] Dropout-correcting codes

[Patrick Giagnocavo]

der Mouse wrote:
> Thanks to struggling with a machine whose only working channel to the
> outside world is a serial line, I find myself caring about encoding
> data such that dropouts can be recovered.  However, after some
> searching, it appears that most of the existing theory for erasure
> codes assumes that you know where the erasures are.  In many cases,
> such as Internet packet streams or big RAID arrays, this is valid (you
> know you didn't get packet #N; you know you got a read error from drive
> M), but in my case it's not.  Can anyone point me at any theory for
> constructing codes to detect and deal with dropouts when you *don't*
> know the dropouts' locations?
> 

There is Reed-Solomon error correction, used on DVDs etc.

http://en.wikipedia.org/wiki/Reed-Solomon_error_correction


follow the "see also" section, including this:

http://en.wikipedia.org/wiki/Low-density_parity-check_code

--Patrick