New features of WinSnoori 1.32
New organization of the editing tools
Addition of white noise ("Editing tools" in the "Edit" menu)
OLA filtering along several time x frequency trajectories ("OLA filtering"
in the "Editing tools")
Changing the highest frequency of the spectrogram (In "Options" from the
Possibility of changing the voicing of F0 points by moving them point by
point (CTRL + left button of the mouse)
Recording at 22050 Hz ("File" menu)
Resampling of signals ("Edit" menu)
Computation of F0 for a corpus
Long filenames can now be used (in My documents for
The F0 computation does not stop one signal window
(4s) before the end of the speech file anymore.
Reassignment of energy in spectrogram computation (Spectro/Reassigned)
Spectrogram computed with critical band analysis (Spectro/Reassigned) menu.
Spectral slices computed with critical band analysis and MFCC cepstral smoothing. This can be useful to investigate analysis parameters used in automatic speech recognition.
- Spectral slices can be exported to the clipboard. Images can then be imported in Paint (Paste) and edited or saved in any image format.
- Remove DC (direct current) component of files (in the file menu). This can be useful when speech was recorded with a poor quality microphone.
- Graphical interface of the Klatt synthesizer. Here are the new functions:
- It is now possible to save the synthesized signal (Save synthesized speech). In addition the default synthetic file is kept in c:\Program Files\Wsno\temp.
- It is possible to remove first or last point of trajectories (Edit menu).
- A copy synthesis function has been added. It adjusts formant amplitudes for the parallel branch and set the frication amplitude.
- The LF source has been added in the Klatt synthesizer. The copy synthesis works but probably needs further adaptations to give better results.
- Automatic formant tracking has been added in the graphical interface of the Klatt synthesizer (Copy synthesis menu). This enables quasi automatic copy synthesis for the formant synthesizer of Klatt.
- Displaying peaks of spectra smoothed with the true envelope algorithm has been added in Formant menu. The true envelope algorithm (S. Imai and Y. Abe, "Spectral envelope extraction by improved cepstral method", Trans. IECE, Vol. J62-A(4), pp. 217-223, 1979) is an iterative version of cepstral smoothing that approximates peaks of spectra (i.e. mainly harmonics). The resulting spectrum is thus more relevant.
- Implementation of the PLP algorithm (Perceptually based Linear Prediction). This algorithm can be accessed from the spectral slice window.
- Addition of three extra formants in the Klatt synthesizer. These formants can be used to represent either nasal formants or other spectral peaks, in the transient part of bursts for instance.
- A new installation program has been created. It uses Inno Setup which is a free Windows intaller. The corresponding WinSnoori version is 1.34.
- Improved MEL scale that gives a better frequency resolution between 0 and 3000 Hz. Is available in the spectral slice computation.