Just suggesting a small optimization:
Each variance value in *.pdf files is sent to HTS_finv() function.
Why not precalculate the inverted values into the *.pdf files and get rid
of the HTS_finv() altogether? Of course this would require changes to the
HTK's and/or HTS demo build scripts' *.pdf creation code.
Getting rid of HTS_finv() enables certain other precalculations and code
optimizations, but no point discussing them here if this considered a bad
idea.
br,
Nicholas
__________ Information from ESET NOD32 Antivirus, version of virus signature database 4767 (20100113) __________
The message was checked by ESET NOD32 Antivirus.
http://www.eset.com