You might like to take a look at the GCG utility
This predicts the coding/non-coding region of
a stretch of DNA without any knowledge of codon-usage
for the given species.
As I understand it, it effectively makes its own
codon frequency table in all three frames and analyses
non-random usage of codons vs. random usage.
It then plots out the probability of coding (non-random)
and non-coding (random) against position such that you
get a graphical prediction of intron/exon regions.
If you have a codon-table file .. you can use 'frames'
Hope this helps,