What is VoiceSauce
It is a software for analyzing acoustic parameters of audio
files.
Specifically, it can analyze voicing source-related parameters,
such as the amplitude of harmonics, spectral tilts, noise level
(Harmonic-to-noise ratio, Cepstral Peak Prominence, Subharmonic to
Harmonic Ratio), Energy, Strength of Excitation (the amplitude of
voicing).
It has the advantage of comparing between different algorithms.
For F0, it contains the measures of STRAIGHT, SNACK, and Praat. For
formants, it has SNACK and Praat. You can also develop your own
algorithm and manually enter values for specific parameter.
VoiceSauce output a measure for every 1 millisecond of the audio.
You can either output all the values or divide the audio into several
equal-timed intervals and output the mean of each interval.
Citation:
- Y.-L. Shue (2010), The voice source in speech production: Data,
analysis and models. UCLA dissertation.
- Shue, Y.-L., P. Keating , C. Vicenik, K. Yu (2011) VoiceSauce: A
program for voice analysis, Proceedings of the ICPhS XVII,
1846-1849.
To install
VoiceSauce
To access the full features of VoiceSauce, please download
VoiceSauce.exe (Windows users) or Matlab (Mac users) following the
instructions below. If you prefer not downloading softwares to your
computer, you can go to Matlab online to access VoiceSauce with features
limited to measuring F0, noise level, Energy, Amplitude of voicing.
Windows users
- Go to http://www.phonetics.ucla.edu/voicesauce/. Under
“Compiled Matlab executables - Windows 7/10”, click on Matlab Component
Runtime and install the installer.exe
installpage
- To find out whether your computer has 32-bit or 64-bit system, go to
“Start” → “Settings” → “About”. On the main page you will see “System
type.”
systemtype
- Download VoiceSauce_bin.zip, unzip the .zip folder, and click on
VoiceSauce.exe to run the program.
wininterface
Mac users
- Go to http://www.phonetics.ucla.edu/voicesauce/. Under “Matlab
m-code”, click on “VoiceSauce.zip” to download it. After download, unzip
the VoiceSauce into a regular folder.
macinstallpage
- Install Matlab so that we can open VoiceSauce in Matlab:
- Go to Matlab support at UW: here.
Click “Sign in to get started”. Log in with your UW username and
password.
- New Users: Create a MathWorks Account. After entering your
information, you will be sent an email to verify this account. Log in
with your newly created MathWorks Account to download the software.
- Returning users: Log in with your MathWorks Account information to
download the software.
- After logging into your Matlab account, select “Install MATLAB”.
Select “R2022b”, click on “Download for macOS”. Open
“matlab_R2022b_maci64.dmg”. The installation will start.
- At the step of “Select products”, select “MATLAB” and “Signal
Processing Toolbox”. Then proceed to the end of “Begin install”.
toolbox
- Open Matlab. Click on the rightmost icon at the address bar “Browse
for folder”. Navigate to the location where the VoiceSauce folder was
stored. Select the VoiceSauce folder and click “Select folder”.
browse
- On the left, find “Current folder” panel, find “VoiceSauce.m” and
double left-click on the file.
currentfolder
- The scripts show up in the Editor in the main panel. Under the tab
of “EDITOR”, click on “Run”. The interface of VoiceSauce shows up.
runscript
interface
How to use
VoiceSauce
Prepare your .wav and
.Textgrid files:
- Put your audio file and TextGrid files in the same folder.
- The .Textgrid file should have the same name as its corresponding
.wav file. Avoid any special characters (e.g. IPA symbols /ʔ, ə, ɯ/,
letters with diacritics /ä, ã/).
- You can assign a different letter to the special characters and
create a code sheet to keep a record of their correspondence.
- A sample folder with audio and Textgrid files can be found here.
Settings in
VoiceSauce
- Click on “Setting”. Under “Common”, change “Not a number label” as
“NaN”.
- Other parameters that you can adjust:
- F0: Max/Min F0
- Formants: Praat Max formant freq; Number of formants;
- Textgrid: Tier numbers (i.e. if you have multiple tiers in your
Textgrid file, which tier you’d like to analyze.)
Parameter
estimation
- Click on “Parameter Estimation”
- Under “Input (*.wav) directory”, click “Browse”, navigate to the
folder where your .wav and .Textgrid files are stored, and click “select
folder”. You will see a list of the files in that folder in the upper
panel of the window.
- If you’d like to save the output .mat files in the same folder with
the sound, check “Save *.mat files with *.wav files”. If you’d like a
different location, navigate to that location under “Output (*.mat)
directory”.
- Click “Parameter Selection” and select the parameters you want to
estimate. If you do not have Praat installed, deselect F0 (Praat) and
Formants (Praat).
parametereselection
- Deselect “Process using 16kHz sampling rate” if that is not what you
want. Check “Use .Textgrid segmentation information if available”.
- Click “Start!” to start the parameter estimation.
parameterestimation
- After seeing the message “Processing complete”, close the window.
estimationfinish
Output to Text
- Click “Output to Text”
- Input .mat directory: where you saved the .mat output
- Input .Textgrid directory: where you saved the .Textgrid files
- Include EGG data: if you have EGG data that you want to include,
navigate to the place where you saved .egg files
- Output .txt directory: where you want to save the .txt output
- Sub-segments:
- No sub-segments: output all the data that are measured every 1
millisecond
- Use sub-segments: the number of mean intervals you want to have for
each target sound. E.g. if you enter “3”, it will divide each target
sound into three equal-timed intervals and calculate the mean value of
each parameter for each interval.
- Parameters: select the ones that you want to include in the .txt
output
- Output Options:
- Single file; You can customize the name of the output file under
“Output file”
- An example output setting where I output H1*-H2*, CPP, Energy,
HNR05, Formants, SoE and output only the mean for each target sound:
outputtext
- After finishing the input, click on “Start!”
Analyze the output in
Excel
Open the output.txt
in Excel
- Go to “Data” tab → “From Text/CSV” → Navigate to the location where
output.txt is saved and select “output.txt”
- Split the “Label” by go to “Data” tab → “Delimited” → Other “-” →
“Next” → “Finish”