Using programs for 11-752

This pages gives instructions for how to use the the programs and database that are part of 11-752: Speech: Phonetics, prosody, perception and synthesis given in the Spring semester in LTI by Maxine Eskenzai and Alan W Black.

Available Machines

These instructions are primarily for the cluster machines in WeH 5336. The programs used in the course are set up to run under Linux on CS facilitized machines. It may be possible to use other machines too, if you decide to do that it is your responsibility to ensure that the software works on them.

Accessing the programs and data

The data, libraries and programs are in

/afs/cs.cmu.edu/academic/class/11752-s00/

To use them add /afs/cs.cmu.edu/academic/class/11752-s00/bin to your path.

How you add this to you path depends on your shell (the program you type commands to). To determine which shell you use

finger

The seocnd line, second column will tell you. If you are using bash, add the following lines to the end of the file .bashrc in your home directory.

export SPPPDIR=/afs/cs.cmu.edu/academic/class/11752-s00
export PATH=$SPPPDIR/bin:$PATH

If you are using csh or tcsh then add the following lines to the of the file .cshrc in your home directory

setenv SPPPDIR /afs/cs.cmu.edu/academic/class/11752-s00
setenv PATH $SPPPDIR/bin:$PATH

(For those of you who care, the course directory as pre-pended to the path list becuase there is an old version of Festival in /usr/local/bin).

You can either logout and in again to have these variables take effect or you type them at the prompt (in the case of tcsh you must also type rehash for them to take effect).

On the lti machines you may want to also add the following line to you .xinitrc in order for delete/backspace

xmodmap -e "keycode 22 = BackSpace"

Using Audio

Using your headset, plug the headset and mircrophone leads into the back of the machines. Where and how the ports are labelled on the back of the machine varies. The microphone socket are usually labelled with a picture of a microphone while the headset output usually has an incomprehensible labelling. Note there are often two output sockets, one for the headset and one for the line out. Both may work (one has a pre-amp), but the headset socket is probably better. There is a program xmixer (in the course bin directory) which allows you to modify the various options controling the audio. In general you want volume and pcm to be high (though you may also want to adjust mic and igain too. The exact best seettings of these depend on the particular machine. Not that some of the headsets have an on/off switch from the microphone

Running the programs

To sets things are working type

say_display_sg

You'll have 5 seconds to speak, say She had your dark suit in greasy washwater all year. (or something similar). After recording the display/label program emulabel will display the signal with a spectragram. You should be able to scroll through it and play parts by pressing the play button. You may optionally give an argument to say_display_sg stating the number of seconds to record.

Other programs you will use:

say_diplay_f0 [SECONDS]
record and display a spectragram and F0 contour. If SECONDS is unspecified it defaults to 5.
na_record -time TIME -o FILENAME.wav
record into FILENAME.wav for given time.
display_sg FILENAME.wav
display the named waveform file and spectragram in emulabel.
display_f0 FILENAME.wav
display the named waveform file and F0 contour in emulabel.
na_play FILENAME.wav
play a waveform.
festival
the festival speech synthesis system. On the Unix command line
- festival enter interactive mode.
- festival --tts FILENAME.txt
  Speak the given filename
- festival -b FILENAME.scm
  oad the given Scheme file and exit
- festival FILENAME.scm
  load the given Scheme file and enter interactive mode
At the festival> interactive prompt
- (SayText "Hello world.")
  speak a string.
- (tts "filename.txt" nil)
  speak a file.
Note in interactive mode standard command line editing works.

Using other machines

All the software we use is available free for at least research use. If you want to install this software on some other machine please speak to Alan W Black.

This page is maintained by Alan W Black awb@cs.cmu.edu