This is quick install instructions for UNIX / Linux systems.
============================================================

Untar the whole archive: 

tar xfzv speakerid-*.tar.gz 

If you don't have fftw and|or libsndfile already installed to your
system, you should untar their respective packages and install them, 
following their respective documentation. If for some reason their
header files are needed, then those can be found on: speakerid/include

Go to source directory: 

cd speakerid/src

Compile the whole thing in Linux:

make

Compile the whole thing in Solaris:

make -f Makefile.solaris  

Now you have binaries: adduser, identifyuser, removeuser and 
                       featextract

With adduser, you can add users to speaker database. You have to 
supply as parameters to it wav-file, path to speaker database (it can 
exist or not), unique ID numeber and path to speaker models (This is not 
mandatory anymore. If parameter is not supplied, then current working 
directory is used).

With identifyuser you can identify users from speaker database. You have to
supply as parameters to it wav-file and path to speaker database (it can 
exist or not).

With featextract you can compute cepstrum feature vectors from 
audio file. You have to supply as parameters to it wav-file and path to 
speaker database (it can exist or not).

Both identifyuser, adduser and featextract use the same getopt_long interface 
for parsing commandline options. Removeuser still uses the standard getopt 
interface, thus their UI differs quite a much. 

You can now set all the parameters in commandline. 

With removeuser you can remove users from speaker database file. You need to
supply as parameters only unique ID and path to speaker database.

These programs demonstrate how to use speakerid API. 

So now you are already ready to try it out. 

Last changed: Fri Jun 14 23:16:59 EET DST 2002