Now, we are going to compile the file Lab0_FE.C we have just completed into the executable FETool, a program that prints out feature vectors for utterances. The traditional thing to do at this point would be to create a Makefile describing which files need to be compiled and linked together to form this executable. However, at IBM we have developed a program named smk that uses artificial intelligence, fuzzy logic, and support vector machines to automatically decide which files need to be linked together, and this program has been made available to you, the student, through the magic of a license agreement.
To compile the program FETool, simply type
smk FETool |
Now, try running FETool with no arguments. It will print out usage information, including a description of all of its command-line arguments. Mastery of the user interfaces of speech recognition tools is notoriously hard to come by, so instead of attempting to foist this knowledge upon you, we will provide shell scripts that supply the appropriate command-line arguments for each experiment that you will need to run. Nevertheless, it is hoped that some of you, by examining these scripts and possibly some source code, will some day be able to figure out how to devise command-line arguments by yourselves.
For this lab, run the script lab0p4.sh by typing
lab0p4.sh |
Save the output of this run in the file p4.out by typing
lab0p4.sh > p4.out |
(To give a little more detail of what FETool does, to use it you must specify where a set of waveform files are located and what signal processing modules to apply and in what order. It then iterates through the waveform files, applies each of the specified signal processing modules to the waveform in turn, and then prints out the resulting feature values.)