Hi,

I've been using Praat for a long time on Windows without any problems,
but encountered a little problem when I recently downloaded and
installed the latest version on Praat on a Mac.  It's running MacOS
X.3, and the problem is that I can't get Praat to play any sound out
on this Mac.  Other applications on this machine will produce sound,
no problem.  It's just Praat that won't produce any audio.  Has anyone
else encountered this problem running Praat on the Mac?

Thanks,
Eric

Hi,

How can I calculate SNR using praat?
Is it possible to add noise to a clean file to get a specific SNR?
can two sounds be added (sound1)+(sound2)?

thanks

Hello!

   I had a query regarding the analysis window portion of the PRAAT sound editor.
   Presently, one can view 5 analysis countors: spectrogram, intensity, pitch,
formants and pulses.
   Is it possible to add some other analysis to this menu eg: Sub-Band energy,
ZCR etc through scripts?
   If not, then what sections of the source code does one need to study and
modify so as to do this?

   I use PRAAT primarily as an analysis tool for understanding my  experiments.
Therefore,  being able to add a "generic analysis" to  the menu would be very
useful.

   Thanks

   Manohar
   DAPLAB, IITB.


---------------------------------
Talk is cheap. Use Yahoo! Messenger to make PC-to-Phone calls.  Great rates
starting at 1¢/min.

[Non-text portions of this message have been removed]

Dear Dr. Boersma,

I know this is a really basic question. But I still don't understand.

This quoation is from a message (#1501) you posted previously.

Indeed, the FFT size expressed in number of samples is an utterly
uninteresting and irrelevant setting, although some programs use it
and many authors report it, perhaps stemming from the 80s when the
only sampling frequency was 10 kHz. The only relevant setting is the
duration of the analysis window. This is what determines in a
sampling-frequency-independent way whether you have a broadband or a
narrowband spectrogram. Try to think in physical terms instead of
computer implementation terms.

In a Praat Manual, I found this:
To get a broadband spectrogram, set window lenghth at 5 ms
(bandwidth 260Hz) and to get narrow a narrowband spectrogram, set it
at 30ms (bandwidth 43Hz)

What I don't understand is how these bandwidths are computed without
knowing a certain sampling frequency. Bandwidths vary depending on
sampling frequency. For a given window length, e.g. window length of
256 points,
10kHz: 10,000/256= 39Hz, and 22.05 kHz: 22050/256=86Hz.
My question is where 260Hz and 43Hz are from without a certain
sampling frequency. And why specifically, 5ms for a broadband
spectrogram and 30ms for a narrow-band spectrogram? Are these
numbers kind of "ideal" length for broad and narrow-band spectrogram
without regard to sampling frequecny?

This is a long mail and my question may not be clear. Sorry. But
still don't understand "why sampling frequency doesn't have to cared
for" Because Praat resamples original signals to a certain frequency
for narrow and broadband spectrogram dispaly?

Thank you for reading this long mail.

Dear straub245

How about thinking this way:

(1) 256-point FFT returns 256 frequency components below the
     sampling frequency (eg, 10 kHz).
     In other words, it divides the 0-10kHz frequecy range
     into 256.
     This is relevant to the frequency resolution.

(2) Suppose the analysis window is fixed (eg, 25.6 ms).

(3) Suppose you double the sampling frequency.

(4) Then you get 512 points in the window, get 512-point FFT,
     get 512 frequency components below 20 kHz, resulting in
     the same frequency resolution.

Thus, the sampling frequency does not affect the frequency
resolution.

On the other hand, if you keep the same sampling frequency
and double the window length, you get 512 frequency
components below 10 kHz.

In short, the window length is the parameter that can
determines the frequency resolution regardless to the
sampling frequency. If you would want to use FFT size to
determine the frequency resolution, you would need to set
the sampling frequency, too.

Sorry, but the exact formula that determines the bandwidth
in Praat is beyond my capability to explain.

Hope this helps a bit.

Masahiko Komatsu

I would like to output the data from a Spectrum plot to a text file.
This script works:

To Spectrum... yes
Edit
To Ltas... 4.882813
To Matrix
n$ = "C:\TDT System II\play\" + name$ + ".txt"
Write to text file... 'n$'

but gives a plot that is lower resolution than the Spectrum plot.  I
need to find the peaks in the Spectrum plot, hence the need for
similar resolution.  How can I do this?

Hi,
   I want to do voice conversion just by modify pitch contour.
Now I get the pitch contour and simply modified it. How can I generate
a new sound which has the modified pitch contour?
   Can you give me some keywords or paper list? I just want to know the
detail, now how to use praat.

Thanks in advance.

Best Regards,
Charles

Hello Praat'ers

I hope that someone can help me with the following question. How can I
make a 'realistic' cepstrum from a voiced signal? I have tried to do
it like this:
- Record mono Sound... (with different sample rates - the lower the
sample rate, the faster the cepstral analysis)
- To Spectrum (fft)
- To Cepstrum
- Draw... 0 0 0 0 Yes
This always give me an amplitude-quefrency graph (= cepstrum), but
this graph doesn't look like a cepstrum at all: it shows a straight
line from left to right (?).
Thank you, Youri

At 02:06 +0000 4-5-06, straub245 wrote:
>What I don't understand is how these bandwidths are computed without
>knowing a certain sampling frequency. Bandwidths vary depending on
>sampling frequency. For a given window length, e.g. window length of
>256 points,
>10kHz: 10,000/256= 39Hz, and 22.05 kHz: 22050/256=86Hz.

Those numbers are the frequency spacings of the resulting FFT. These numbers are
not the bandwidths.

If Praat takes a window with "window length" 5 milliseconds from a sound sampled
at 22,050 Hertz, then Praat actually takes 10 milliseconds because of the
Gaussian window shape. This would be 221 samples under the bell curve. With the
standard settings (a complicated computation) Praat will add zeroes to turn it
into 512 samples, giving 257 frequency bins, each of which is 43 Hz wide.

So in this example, the frequency is measured in 43 Hz steps (i.e. the bin width
is 43 Hz), but the bandwidth (frequency resolution) is still 260 Hz.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

At 14:51 +0000 3-5-06, neediss wrote:
>  I want to do voice conversion just by modify pitch contour.
>Now I get the pitch contour and simply modified it. How can I generate
>a new sound which has the modified pitch contour?

read the Intro, part 8.1.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

hi,

In advance, I am sorry if I ask something very trivial.
I have a frustrating problem:

#ok
Create FormantTier... filter 0 0.5
Add point... 0.0  350 50


"You can use numeric variables and formulas in arguments to commands:"
(http://www.fon.hum.uva.nl/praat/manual/Scripting_5_2__Formulas.html)
so i tried:

#wrong
variable = 350
Create FormantTier... filter 0 0.5
Add point... 0.0  variable 50


this code crashes with:
"Number of formant-bandwidth pairs must be at least 1."

If the variable is in a later argument, than it is simply skipped.
I am not really sure whether I have a syntactic error, type mismatch
(arguments should be pairs-of-values?)or ... ?

i looked around in this list (and on the net), but i couldnt find
anything that is related.

thank you very much for your help.

gabotan

At 05:25 +0000 6-5-06, gab0tan wrote:
> "You can use numeric variables and formulas in arguments to commands:"
> Add point... 0.0  variable 50
> "Number of formant-bandwidth pairs must be at least 1."

will, this argument is a list of numbers rather than one number, and this is a
counterexample to the rule that one can use variables and formulas. So you would
have to do:

   Add point... 0.0 'variable' 50

By the way, the manual is easier to consult under Help.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

Hey all,
I added a couple of buttons to the dynamic menu and could not remove
them. Have tried the "button" and also reinstallation. Is it possible
to restore the original layout?
Thanks a lot!

At 01:29 +0000 7-5-06, lb9121 wrote:
>I added a couple of buttons to the dynamic menu and could not remove
>them. Have tried the "button" and also reinstallation. Is it possible
>to restore the original layout?

Go to "buttons" and look them up. They are under the object type that you
specified when installing the button. For instance, if the buttons are visible
if one selects a Sound, select the N-Z actions and scroll down.

You can also just remove the Buttons file (see the manual).
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

Hi all,

Can anyone explain how the available pitch smoothing function proceeds?

Thanks in advance,
Francisco

Dear all,

Is there a way to calculate the noise component in running speech either
periodic or aperiodic?

The Harmonics-to-Noise ratio seems to be meaningful for periodic signal
only.
Any suggestions (and/or pieces of PRAAT scripts dealing with it) would be
appreciated.


Best regards
Olga


----------------------------------------------
Olga Gordeeva
Speech Science Research Centre
Queen Margaret University College
Edinburgh UK

>From: Paul Boersma <paul.boersma@...>
>To: Jasmin Menzel <menz2801@...>, praat-users@...
>Subject: Re: [praat-users] reducing dB
>Date: Wed, 4 Jan 2006 14:34:48 +0100
>
>At 01:30 +0100 4-1-06, Jasmin Menzel wrote:
> >I'm wondering, if it's possible to reduce the dB value for specific
> >frequencies.
>
>yes. If you want to reduce, for instance, the intensity of all frequencies
>between 3000 and 4000 Hz by 17 dB, this means that you want to multiply the
>sound pressure in that region by a factor of 10^(-17/20) = 0.141. That
>reads as: 10 to the power minus 17/20. You can use Praat's calculator to
>compute this value.
>
>You can then use Filter(formula) as follows:
>
>if x>3000 or x<4000 then self*0.141 else self fi
>
>This may do what you want. However, there is a disadvantage, in that the
>filter is quite abrupt at its edges (3000 and 4000 Hz), which may cause
>"ringing", i.e. sinewave-like phenomena at the beginning and end of the
>sound. So the edges of the filter should be smoother, for instance going
>from 1 to 0.141 within a band of 100 Hz:
>
>self * if x<2950 then 1 else if x<3050 then (0.141*(x-2950)+1*(3050-x))/100
>else if x<3950 then 0.141 else if x<4050 then
>(1*(x-3950)+0.141*(4050-x))/100 else 1 fi fi fi fi
>
>Slightly cumbersome, isn't it? But it does work, as you can see when
>comparing the spectra. This formula represents a filter whose spectrum
>consists of five lines and looks like a bucket. If your email program has
>all letters equally wide, it will look like this:
>
>___       ___
>    \_____/
>
>
>--
>
>Paul Boersma
>Institute of Phonetic Sciences, University of Amsterdam
>Herengracht 338, 1016CG Amsterdam, The Netherlands
>http://www.fon.hum.uva.nl/paul/
>phone +31-20-5252385

I can't download new version 4.4.20 for mac. Error 95. Thanks

At 10:59 +0000 8-5-06, Olga Gordeeva wrote:
>Is there a way to calculate the noise component in running speech either
>periodic or aperiodic?
>
>The Harmonics-to-Noise ratio seems to be meaningful for periodic signal
>only.

since "noise" in speech is defined as anything not periodic, the
"signal-to-noise ratio" for speech automatically means the ratio of harmonic
spectral components to nonharmonic spectral components.

With our knowledge of what the glottal source does, we can usually measure some
of the irregularities that lead to noise, such as "jitter" and "shimmer".
Regrettably, a subtraction of "jitter" and "shimmer" from the sound, presumably
giving the remaining "additive noise", is not very well possible (to my
knowledge). But I don't know anyway if this is what you are aiming at.

The problem with asking for the noise component in aperiodic sound is that I
would not know in general what the nonnoise component in aperiodic sound could
be, i.e. what other kind of "signal" (other than periodic) you could want to
find in speech sound. Perhaps you have some suggestions.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

>The problem with asking for the noise component in aperiodic sound is that
>I would not know in general what the nonnoise component in aperiodic sound
>could be, i.e. what other kind
>of "signal" (other than periodic) you could want to find in speech sound.
>Perhaps you have some >suggestions.

I am comparing BUS/BUZZ tokens of which the BUS tokens contain a relatively
long aspirated VC transition ([h]-like preaspiration). I am trying to
pinpoint the differences in transitions acoustically and in timing.

A signal-to-Noise Ratio throughout the two vowels including transitions
would be a good indication of the BUS/BUZZ differences in pre-aspiration.
However the [h]-like VC transition in BUS often lacks periodicity and H-to-N
ratio gives back -200 db values, which I don't think are meaningful. So the
ammount of noise is 100% from that point, and I can not use HTN ratio any
more. Is there a way to solve this? I assume that the ammount of noise is
known in the periodic component?

Is there a way to get the same consistent noise measure throughout such
vowel+[h]-like transition that could apply to both periodic and aperiodic
parts of this VC transition?

Many thanks for any suggestions

Olga Gordeeva





>From: Paul Boersma <paul.boersma@...>
>To: "Olga Gordeeva" <olgagordeeva@...>,
>praat-users@...
>Subject: Re: [praat-users] noise component in running speech
>Date: Tue, 9 May 2006 16:59:41 +0200
>
>At 10:59 +0000 8-5-06, Olga Gordeeva wrote:
> >Is there a way to calculate the noise component in running speech either
> >periodic or aperiodic?
> >
> >The Harmonics-to-Noise ratio seems to be meaningful for periodic signal
> >only.
>
>since "noise" in speech is defined as anything not periodic, the
>"signal-to-noise ratio" for speech automatically means the ratio of
>harmonic spectral components to nonharmonic spectral components.
>
>With our knowledge of what the glottal source does, we can usually measure
>some of the irregularities that lead to noise, such as "jitter" and
>"shimmer". Regrettably, a subtraction of "jitter" and "shimmer" from the
>sound, presumably giving the remaining "additive noise", is not very well
>possible (to my knowledge). But I don't know anyway if this is what you are
>aiming at.
>
>The problem with asking for the noise component in aperiodic sound is that
>I would not know in general what the nonnoise component in aperiodic sound
>could be, i.e. what other kind of "signal" (other than periodic) you could
>want to find in speech sound. Perhaps you have some suggestions.
>--
>
>Paul Boersma
>Institute of Phonetic Sciences, University of Amsterdam
>Herengracht 338, 1016CG Amsterdam, The Netherlands
>http://www.fon.hum.uva.nl/paul/
>phone +31-20-5252385

At 15:44 +0000 9-5-06, Olga Gordeeva wrote:
> A signal-to-Noise Ratio throughout the two vowels including transitions would
be a good indication of the BUS/BUZZ differences in pre-aspiration. However the
[h]-like VC transition in BUS often lacks periodicity and H-to-N ratio gives
back -200 db values, which I don't think are meaningful.

yes. This is because there is no periodicity. By definition (i.e. by the
definition of noise), all of the sound is noise, and the HNR is zero (= minus
infinity dB).

> So the ammount of noise is 100% from that point, and I can not use HTN ratio
any more. Is there a way to solve this?

As I explained, the amount of noise is indeed 100%. What do you mean by "solve"?
What part of the aperiodic sound would you instead want to regard as non-noise?
In aperiodic sounds, there *is* no non-noisy part (by definition).

> I assume that the ammount of noise is known in the periodic component?

You mean in the more-or-less periodic part of the sound? Yes, it's the
non-periodic component of the more-or-less periodic part of the sound.

By the way, one can make Praat hallucinate that there is some periodicity in
voiceless sounds. From what you tell me, I gather that you use the "Voice
report" to tell you the degree of periodicity (HNR), but it will do so only for
the parts that are considered voiced in the pitch analysis (because that is what
voice reports are about). If you extract the voiceless parts that you are
interested in to the Objects window, then choose "To HNR", Praat will try to
find the flimsiest excuse for regarding some component of the sound as periodic,
so that it will typically yield HNR values (even for really "white" noise)
between -10 dB (= 90% noise) and 0 dB (= 50% noise). Such measurements tend to
distinguish between Dutch "s" and "z", for instance, although both may end up
being analysed as voiceless in the pitch analysis. However, these numbers like
-5 dB and so are not very meaningful in themselves; although they discriminate
between fricatives, they represent a hallucination that slightly depends on the
"minimum pitch" setting, for instance.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

I was wondering if you could help me with this:

I need to remove single words from my data due to Data Protection
issues but I would like to be able to do this without leaving a blank.
Ideally I would be left with some sort of schwa sound so I can still
hear the pitch etc. of the removed words.
I've noticed in synthesis you can change the sound to a hum which is
ideal but I'm not sure how to execute this!

I am very new to Praat so any advice someone could give me on this
would be very gratefully received

--- In raat-users@...">praat-users@..., Paul Boersma
<aul.boersma@u">paul.boersma@u...> wrote:
>
> At 16:54 +0000 16-2-06, kotzbroeckel wrote:
> > And how can I create another line in the endtext of the experiment?
>
> That does not work at the moment. If you try the following:
>
> endText = "The experiment
> has finished."
>
> then the text will still appear a single line, although this may change in the
future.




Has there been any improvement with this? I am trying to use a second line or
shrink the fonts for my text to fit in the screen, but I don't find a way to do
it!




--- In praat-users@..., "psibre" <steiner@c...> wrote:
>
> 3) Try inserting 'newline$' into your output string



I tried using this 'newline$' but I seem not to be doing it correctly. If I
wanted to have a new line in the example used above (endText = "The experiment
has finished."), how should I use this 'newline$'?

Thanks for your time,

Spyros




[Non-text portions of this message have been removed]

Hello Suzy

that was an interesting question! I think we may also may need to use
this tool

the following script should do the trick

you need to create a TextGrid for your sound file with a special
label for the words you want to censure - the default tier is 2 and
the default label is "hum" but of course you can change those

the result sounds a bit loud for the hum compared to the normal voice
- maybe it would be a good idea to reduce the volume of the hummed
portions.

Hoping this is useful - let me know if you have any problems.




   ----------


best wishes

daniel



------------------------------------------------------------------------
-------------------------
    Equipe Prosodie et Représentation Formelle du Langage - salle A485
                 CNRS Laboratoire Parole et Langage (UMR 6057)
                    Université de Provence 29 avenue Schuman
                         13621 Aix-en-Provence cedex 1, France
    tel: +33-4-4295-3628    secr: +33-4-4295-3634   fax: +33-4-4295-3788
------------------------------------------------------------------------
--------------------------
Speech Prosody list      http://mailup.univ-mrs.fr/wws/info/speech-
prosody
Praat-users list                           http://groups.yahoo.com/
group/praat-users
Speech Prosody 2006 May 2-5    http://www.ias.et.tu-dresden.de/sp2006
------------------------------------------------------------------------
--------------------------




[Non-text portions of this message have been removed]

At 09:53 +0000 12-5-06, suzy_orr wrote:
>I've noticed in synthesis you can change the sound to a hum which is
>ideal but I'm not sure how to execute this!

It requires same manual work:

1. Select the word in the sound window (and remember the times).
2. Extract the selected word as a Sound to the Objects list (File menu).
3. Choose Get Intensity (dB) from the Query menu, and remember the result.
4. Choose To Pitch and click OK.
5. Choose To Sound (hum) and listen to the resulting sound;
    it has a straight intensity curve, which is not yet what you want.
6. Select again your original extracted word in the Objects list.
7. Choose To Intensity and click OK.
8. Choose Down To IntensityTier.
9. Select the hummed sound plus the IntensityTier.
10. Choose Multiply, click OK, and listen to the resulting sound;
    it has the correct intensity curve, but will still be to loud.
11. Select the almost ready sound in the Objects list,
    and choose Scale Intensity from the Modify menu;
    in the Scale Intensity window,
    type the number that you remember from step 3 and click OK;
    you now have the correct hum.
12. In the window with the original sound,
    select the same times as in step 1 (the times may still be there).
13. Choose Cut (Edit menu; the word disappears).
14. Select your schwa and choose Edit.
15. In the sound window, select the whole sound (the schwa),
     and choose Copy (Edit menu).
16. Go back to the window with the original sound and choose Paste (Edit menu).

There may be some irregularities at the edges, so you may do some of the cutting
and copying on zero crossings.

--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385

sorry - as owner of this list I should know by now that we are not
allowed to attach files !!!!

so here is the script:

save it as a text file with the name censured.praat then open it from
praat with Open Praat script...

select a Sound file and a TextGrid file in which you have put special
labels (the default is "hum") on tier (default = 2)

best wishes

daniel

=======================================================

#praat script
scriptName$ = "censured.praat"
version$ = "2006:05:12"
#author: Daniel Hirst
#email: daniel.hirst@...

#purpose: replace parts of a Sound with a hum
#requires: one Sound file and one TextGrid file both selected
# 	 the form allows the definition of the tier containing the
# 	 labels of the portions of the sound which are to be censured
# 	 as well as the name of the labels identifying them.


form select the tier and the labels of the parts of the sound you
want to censure
	 natural Tier 2
	 word target hum
	 natural minF0 65
	 natural maxF0 250
endform

if numberOfSelected("Sound") <> 1 or numberOfSelected("TextGrid")<> 1
	 pause Please select one Sound and one TextGrid
endif

mySound = selected("Sound")
name$ = selected$("Sound")
myTextGrid = selected("TextGrid")
select mySound
finish = Get finishing time
frequency = Get sampling frequency
To Pitch... 0.01 minF0 maxF0
myPitch = selected("Pitch")
select myPitch
To Sound (hum)
Override sampling frequency... frequency
myHum = selected("Sound")
select myPitch
Remove

select myTextGrid
nLabels = Get number of intervals... tier
lastStart = 0
end = 0
for iLabel from 1 to nLabels
	 select myTextGrid
	 label$ = Get label of interval... tier iLabel
	 if label$ = target$
		 start = Get starting point... tier iLabel
		 end = Get end point... tier iLabel
		 if start > lastStart
			 select mySound
			 Extract part... lastStart start Rectangular 1.0 no
		 endif
		 select myHum
		 Extract part... start end Rectangular 1.0 no
		 lastStart = end
	 endif
endfor

if lastStart < finish
	 select mySound
	 select mySound
	 Extract part... lastStart finish Rectangular 1.0 no
endif

select all
minus mySound
minus myTextGrid
minus myHum
Concatenate
Rename... 'name$'-'target$'
myNewSound = selected("Sound")


select all
minus mySound
minus myTextGrid
minus myNewSound
Remove
select myNewSound




------------------------------------------------------------------------
-------------------------
    Equipe Prosodie et Représentation Formelle du Langage - salle A485
                 CNRS Laboratoire Parole et Langage (UMR 6057)
                    Université de Provence 29 avenue Schuman
                         13621 Aix-en-Provence cedex 1, France
    tel: +33-4-4295-3628    secr: +33-4-4295-3634   fax: +33-4-4295-3788
------------------------------------------------------------------------
--------------------------
Speech Prosody list      http://mailup.univ-mrs.fr/wws/info/speech-
prosody
Praat-users list                           http://groups.yahoo.com/
group/praat-users
Speech Prosody 2006 May 2-5    http://www.ias.et.tu-dresden.de/sp2006
------------------------------------------------------------------------
--------------------------

Need some help and am at my wits end with regard to the following:

Basically, what I'm trying to do is:

(1) select a 10ms stretch of sound
(2) Bark-filter the thing
(3) Get it's spectrum (excitation pattern??)
(4) Get its so-called center of gravity, skewness and kurtosis (like
one can do on a normal unfiltered spectral slice in the main object
window - under Query).

Basically the problem is that if I do the FFT directly off the main
sound file and get the power spectrum in return, there is no Bark
filtering option available. There is the "To Excitation" option which
I presume is the Bark-filtered equivalent of the power spectrum (i.e.
Bark on x-axis, some measurement of level of excitation on y-axis),
from which I get a so-called Excitation slice, but then I can't query
this for center of gravity etc. as I can do with a normal spectral
slice...

If I tackle it the other way round i.e. by first selecting the whole
fricative say from the main sound file (by using the extract windowed
selection option) and then Bark filter, I get something called a Bark
filter slice (which I presume is different to an excitation slice but
I'm not sure). Either way I can't seem to query this for center of
gravity etc. either.

So am thoroughly confused.

FYI,
The following forwarded message came through on the through on the TESOL-ITA
list.  A group at the U of Minnesota has developed a very nice beginner info
page for those who use Praat for ESL.
-Aaron Ohlrogge
   aaronohlrogge@...


---------------------------------
Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.

[Non-text portions of this message have been removed]

Hello everyone,

I've been unsuccessful getting Praat to work on Linux PPC (Suse 10.1,
but I got the same problem with the last stable and development Ubuntu).

I compiled it using the linux dynamic defs file against the latest
openmotif. When I try to play a short wav in Praat, what I get is just
pure noise. The sound is otherwise working and I'm able to play the
same file in Audacity, for instance.

Does anyone have a clue about what's going on? I searched the list's
archive but didn't find anything about it.

Thanks in advance for your help,

All the best,

Julien

incorporating Paul's recipe into my script gives the following - and
the result sounds much better!!

thanks both Suzy and Paul - this is a script which will be very
useful to me!

best wishes

daniel



=========================================
#praat script
scriptName$ = "censured.praat"
version$ = "2006:05:12"
#author: Daniel Hirst
#email: daniel.hirst@...

#purpose: replace parts of a Sound with a hum
#requires: one Sound file and one TextGrid file both selected
# 	 the form allows you to specify which tier (default 2) contains the
# 	 labels of the portions of the sound which are to be censured
# 	 as well as the name of the label identifying them (default "hum").


form select the tier and the labels of the parts of the sound you
want to censure
	 natural Tier 2
	 word target hum
	 natural minF0 65
	 natural maxF0 250
endform

if numberOfSelected("Sound") <> 1 or numberOfSelected("TextGrid")<> 1
	 pause Please select one Sound and one TextGrid
endif

mySound = selected("Sound")
name$ = selected$("Sound")
myTextGrid = selected("TextGrid")
select mySound
finish = Get finishing time
frequency = Get sampling frequency
To Pitch... 0.01 minF0 maxF0
myPitch = selected("Pitch")
select myPitch
To Sound (hum)
Override sampling frequency... frequency
myHum = selected("Sound")
select myPitch
Remove

select myTextGrid
nLabels = Get number of intervals... tier
lastStart = 0
end = 0
for iLabel from 1 to nLabels
	 select myTextGrid
	 label$ = Get label of interval... tier iLabel
	 if label$ = target$
		 start = Get starting point... tier iLabel
		 end = Get end point... tier iLabel
		 if start > lastStart
			 select mySound
			 Extract part... lastStart start Rectangular 1.0 no
		 endif
		 select myHum
		 Extract part... start end Rectangular 1.0 no
		 lastStart = end
	 endif
endfor

if lastStart < finish
	 select mySound
	 select mySound
	 Extract part... lastStart finish Rectangular 1.0 no
endif

select all
minus mySound
minus myTextGrid
minus myHum
Concatenate
Rename... 'name$'-'target$'
myNewSound = selected("Sound")


select all
minus mySound
minus myTextGrid
minus myNewSound
Remove
select myNewSound




------------------------------------------------------------------------
-------------------------
    Equipe Prosodie et Représentation Formelle du Langage - salle A485
                 CNRS Laboratoire Parole et Langage (UMR 6057)
                    Université de Provence 29 avenue Schuman
                         13621 Aix-en-Provence cedex 1, France
    tel: +33-4-4295-3628    secr: +33-4-4295-3634   fax: +33-4-4295-3788
------------------------------------------------------------------------
--------------------------
Speech Prosody list      http://mailup.univ-mrs.fr/wws/info/speech-
prosody
Praat-users list                           http://groups.yahoo.com/
group/praat-users
Speech Prosody 2006 May 2-5    http://www.ias.et.tu-dresden.de/sp2006
------------------------------------------------------------------------
--------------------------

Hello,

I'm trying to convert MATLAB algorithms to PRAAT and I'm having some
conceptual difficulties with how to go about doing things in PRAAT.
Below are a few problems which have me stumped:

Q1.  How do I read an array into PRAAT?
I've managed so far to a) read a matrix from a txt file and convert
this matrix to a tableofreal, but with both of these object types
there are only certain functions i can perform on them.  Ideally,
what I would like is to read the txt file and then have the array
from the text file as a variable that I can manipulate in my
script.  Is this possible?

Q2.  If the above is not possible, can I convert the data contained
in a matrix (or a TableofReal) into a variable that I can manipulate
in a script?  I think these are really two ways of asking the same
question.

Q3. How do I find the minimum value in a matrix?  I've seen the minx
and miny functions but neither of these return actual minimum values
from the matrix.  Example, if I have an row array [2 3 4 6 1 4], how
do I return a '1' as the minimum value.

Thank you,

Daniel Little