pfview is a Tcl/Tk program that reads pfiles (i.e. speech feature files) and allows interactive display of the data they contain as grayscale images. Several pfiles may be displayed at once, and it's easy to generate a postscript version of the display. You can also display labels from a separate, but conformal, archive file. Finally, the source waveform will be found and displayed if possible.
The code currently resides in /u/dpwe/projects/pfview/. The executable is pfview.tcl, and you run it with something like:
Once launched, you can switch to other pfiles, create additional displays and generate postscript from the menu. You can also drag the items on the canvas around for greater aesthetic appeal, although they won't let you mess with their time-alignment. The right mouse button brings up a contextual menu that you can use to remove elements from the display.
For a more complex example, the screenshot above was from a viewer launched as below:
~/projects/pfview/pfview.tcl \ pfile=../../../pfile/ras8+d+dd/train-rand.pf \ labfile=train-rand-rasi5+q.ilab \ labtype=ilab \ listfile=../../../list/train-rand.list \ wavfilecmd="echo /u/drspeech/data/aurora/wavfile/train/%u"The pfile argument points to the feature file, and the labfile is the conformal label file. The labtype can be "pfile" or "ilab", depending on the format of the label archive. The listfile gives the utterance ID for each segment in the pfile, and the wavfilecmd is executed, with %u replaced by the utterance ID, to generate the wave filename (possibly followed by start/end times) for waveform display and playback.