Glottalizer 0.9.1 User Documentation

by Nicolas Audibert and Tamás Bőhm, June 26, 2008

GUI layout Run the program Load files Selection, zoom and move
Modify periods Modify pitchmarks Undo/redo actions Set preferences
GUI layout

Figure 1: General layout of Glottalizer's user interface

The bottom panel displays the waveform of the recording to be manipulated. The top panel depicts the model waveform that can be used to guide the transformation (either manually or by copying its pulse pattern); for creating irregular phonation, a model recording that contains irregular phonation can be loaded into this panel.

Both panels have buttons for playing the corresponding signal (either the entire signal, only the selection or only the visible part), for zooming (horizontally and vertically), and for moving the display window.

The manipulation panel also has buttons for playing the original signal before manipulation and switching back to this original signal by setting all the scaling factors to 1, while the model panel has a "Pattern copy" button for applying the characteristics of the region selected in the model panel to the manipulated signal (see below).

The "Other actions" area includes both actions specific to the manipulation panel and common to both panels.

Run the program

On Windows, run glottalizer_win.exe. The program accepts 2 optional parameters (if started from a command prompt):

  • Parameter 1: The name of the file to be loaded into the manipulation panel
  • Parameter 2: The name of the file to be loaded into the model panel

Load files
Manipulation panel

The wave file to be manipulated can be loaded using the "Open file" action in the File menu, the CTRL+O shortcut or the corresponding toolbar icon. There must be a pitchmark file with the same name in the same directory, with the extension .PointProcess (Praat), .pm (raw text file with one pitchmark location per line, expressed in seconds) or .pit (BME-TMIT SpeechLab internal format), supposed to indicate the instants of glottal closures.

Setting the pitchmarks manually can be time-consuming, therefore it is recommended to generate the pitchmark file as a PointProcess using the freely available Praat software as a first step: open the sound with Praat, find pitchmarks and put them into a PointProcess object using Periodicity/To PointProcess..., select the generated PointProcess object in the Praat window and save it using Write/Write to short text file...

When a large number of pitchmarks are at a “suspicious” location, a warning is displayed when opening the file. Note that the algorithm for detecting "suspicious" pitchmarks is very preliminary, visual inspection of pitchmarks remains necessary.

Pitchmarks can be displayed as vertical lines or hidden by checking/unchecking "Show pitchmarks".

Previous manipulations on the same signal saved with the .gain format (i.e. only the scaling factors) can be loaded using the "Load gain" action without reloading the wave file.

Model panel

The model file to be used as a guide for the modification of the manipulated signal can be loaded using the "Load model" action in the File menu. Similarly to the signal to be manipulated, a pitchmark file with the same name and .PointProcess, .pm or .pit extension must exist in the same directory.

Selection, zoom and move

To select a region, left-click the beginning or the end of the desired selection, hold the left button clicked and release it at the other end of the region to select (i.e. the usual Windows selection method). There can be a selection in both panels. A left click cancels the corresponding selection, whatever the edition mode is ("Edit pitchmark" button toggled on or off).

The default display of a signal in either the manipulation or the model panel shows the entire signal with the y-axis scaled according the max and min values of the displayed signal, with extra space on top and bottom for the display of pitchmarks and time marks. However the viewports of x and y-axis can be tuned using the "Zoom and move" panel (figure 2) associated with the corresponding panel.

Figure 2: The "Zoom and move" panel

This panel offers the possibility to zoom in and out (the zooming factor can be parameterized, see Set preferences) on the horizontal and vertical axes, while the "Default" and "All" buttons restore the initial display. On the horizontal axis, the "Sel" button zooms to the selected area, while the "<=" and "=>" arrows can be used to scroll the displayed part of the signal (the scroll step can be parameterized in config.txt).

Modify periods

The transformation method associates a gain value (scaling factor si) with each pitchmark in the manipulation panel, that used for scaling the corresponding period. All the gain values are initially set to 1 (no scaling), and changed to values <1 when the period is attenuated, to values >1 when it is boosted, and to 0 if the period is removed. Gain values can be displayed by checking "Show gain values". All gain values can be reset to 1 (undoing all the manipulations) using the "Revert to original" button.

Attenuating or zeroing a fundamental period also scales down the background noise present during the period, that can decrease the perceived naturalness. In order to avoid this problem, appropriately scaled background noise can be automatically added to attenuated and zeroed cycles by checking "Replace removed periods with background noise". By default, background noise is copied from the end of the signal, but it can be cut out from any other region of the signal by selecting it in the manipulation panel and pushing the "Set background noise to selection" button.

Scaling of individual periods

When a signal is loaded in the manipulation panel and the "Edit pitchmarks" toggle button is disabled (see Modify pitchmarks), individual pitch periods can be scaled, zeroed or reset to its initial value via mouse operations:

  • Periods can be removed (scaled to zero) by right-clicking with the mouse around the corresponding pitch mark. A second right-click brings the period back to its original form (i.e. resets the scaling factor to 1).
  • Periods can be boosted or attenuated by left-clicking around the pitch mark, with the vertical position of the mouse pointer determining the new peak value (and thus the scaling factor). The applied scaling factors are shown above the manipulated waveform.

Pattern copy

The "Pattern copy" push button is enabled only when a region is selected in both panels and there are enough pitchmarks before the selected region in the model panel (to calculate the reference values). Pushing this button automatically calculates scaling factors and applies them to the selected region in the manipulation panel.

Figure 3: Screenshot after copying a pulse pattern (figure 1 shows a screenshot before pattern copy)

Save modifications

The "Save manipulation" action (or the corresponding toolbar item or the CTRL+S shortcut) saves the scaling factors using the last filename given in "Save manipulation as", which is called at the first save of a manipulated signal.

The "Save manipulation as" action in the File menu opens a dialog box where the file name and file format can be specified. The file name is set as a default to the name of the original wave file appended with "_gl". The save format can be set either to .gain (save the scaling factors only) or to .wav (save both the wave file and the scaling factors).

Modify pitchmarks
Add/remove pitchmarks

When the "Edit pitchmarks" mode is turned on (using the toggle button on the bottom right of the window), mouse clicks on axes modify pitchmarks:

  • A left click on a panel where a signal is loaded adds a pitchmark at the clicked position. If a pitchmark already exists close to the clicked position, the pitchmark is not added and a warning is displayed.
  • A right click removes the closest pitchmark on the clicked panel.

Pitchmark-groups are automatically detected and the first and last pitchmarks of a group are displayed in red (figure 4). Periods that correspond to those 2 particular pitchmarks cannot be modified for technical reasons.

Figure 4: Illustration of pitchmark-group boundaries (vertical red lines)

Save modified pitchmarks

The "Save pitchmarks (manipulation)" and "Save pitchmarks (model)" actions in the File menu save the corresponding pitchmarks using the format (.PointProcess, .pm or .pit) specified in the save dialog.

Undo/redo actions

The last manipulation can be undone either by the "Undo" action in the Edit menu, by the corresponding toolbar icon, or by the CTRL+Z shortcut. An undone manipulation can be redone using either the "Redo" action in the Edit menu, the corresponding toolbar icon, or the CTRL+Y shortcut.

Set preferences

Some constants used in period modification, pattern copy and other actions, as well as some used for display, can be tuned by editing the config.txt preferences file. If this file is edited while Glottalizer is running, preferences can be reloaded using the "Reload preferences" action in the Edit menu.

Back to top