Hi

Can anyone tell me why this falls over when I try to get the vocal tract
length?

form  xyz
             word sFile
             word name
             word prtFile
endform

# just to empty the file
filedelete 'prtFile$'

# to read the file

Read from file... 'sFile$'
select Sound 'name$'
Play


# to extract the vocal tract length
# figures are settings

select Sound 'name$'
# don't forget Nyquist - this is 8k sampling therefore 4k limit and 4
formants
# men app 1000 each women slightly less
To LPC (covariance)... 4 0.025 0.005 50
To VocalTract (slice)... 0 0.17 yes
Get Info...
length = Get Vocal tract length
fileappend 'prtFile$' vocalTrackLength => 'length:3' 'newline$

I guess I have written the lines:

Get Info...
length = Get Vocal tract length

wrongly, but I don’t know how to get this item of information.

Stephen



--
Internal Virus Database is out-of-date.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 266.11.16 - Release Date: 24/05/2005



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

I have a question concerning experimentMFC. I would like to have my
subjects choose from 4 files for a total of 160 files (i.e they should
make 40 choices). These 4 files belong together, but should be
presented  in a randomized order.

    Subject 1     Subject 2
    a b c d       a c d b
    e f g h       h f g e

Is this possible in an experimentMFC object?

any help is appreciated,

ruben van de vijver

At 04:09 +0000 1-6-05, evynnt wrote:
>I've written an expMFC that Praat can read, but it won't run it
>because, for some reason, it can't read my sound files for the that
>purpose. The folder I have the sounds in is called "praat" and the
>sounds are .wav files, so that line in the .txt file
>is: "praat\" ".wav".

So you want to use relative path names. That's a good idea, because the
experiment file will then be portable across computers and even across operating
systems. Make sure of two things, though:

1. the directory "praat" should be in the same directory as the experiment file.
2. use "praat/" rather than "praat\", as in the example in 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

At 17:21 +1000 1-6-05, Stephen Choularton wrote:
>Get Info...
>length = Get Vocal tract length

Try this:

info$ = Info
vocalTractLength = extractNumber (info$, "Vocal tract length:")

--

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

I am extracting vocal tract length by throwing short utterance at praat.
Is that reasonable or will it only work on a single vowel?

Stephen

--
Internal Virus Database is out-of-date.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 266.11.16 - Release Date: 24/05/2005



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

Hello,
we tried to write a script in order to save as a txt file the pitch
listing.

In the crucial part of the script:

editor Sound ex1
Pitch listing /Users/...../Desktop/ex1.txt
endeditor

Praat tells that Pitch listing cannot be saved within the editor but
the error comes ot even if I move it out from the editor or if I
recall Pitch listing within the editor and the try to write it to
file after "endeditor"

Thanks for your help.
Marco

Hi

I am having problems with extracting vocal tract length.  I am using this code
to extract from male utterances:

Resample... 10000 50
To LPC (autocorrelation)... 14 0.025 0.005 50
To VocalTract (slice)... 0 0.17 yes
info$ = Info
vocalTractLength = extractNumber(info$, "Vocal tract length:")

The first utterance produced .188 which may well be reasonable but then I got
error messages:

1. copy E:\TIDIGITS_ORIGINAL\TIDIGITS_ADULTS\TIDIGITS\TEST\MAN\TC\1474568
474568A.WAV
         1 file(s) copied.
2. shorten -x -a 1024 1474568A.WAV 1474568A.sph
3. sph_convert.exe 1474568A.sph
Error: LPC_to_VocalTract: VTL could not be calculated.
No object was put into the list.
Script line 32 not performed or completed:
½ To VocalTract (slice)... 0 0.17 yes ╗
Script "C:\CorpMan\getPraat.script" not completed.
Praatcon.exe: command file "getPraat.script 1474568A.wav 1474568A E:\TIDI
IGINAL\TIDIGITS_ADULTS\TIDIGITS\TEST\MAN\TC\1474568A.prt" not completed.

It occasionally produces a figure on other utterances but I get a lot of this
sort of error message.

The next one for this speaker that worked produced .179

A few questions:

Is my script wrong and producing the errors?

If I have phone speech can I upsample from 8k?

Is the difference between .179 and .188 normal  the normal level of accuracy you
get when throwing a whole utterance at Praat?

If I have mixed male/female speech will I get decent results just resampling
everything to 11000?

Thanks.

Stephen

--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 267.5.0 - Release Date: 2/06/2005



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

At 09:37 +0000 2-6-05, legito wrote:
>Pitch listing /Users/...../Desktop/ex1.txt

this does not do what you want. The command "Pitch listing" just writes to the
Info window.

To put the pitch listing into a file, do something like

listing$ = Pitch listing
listing$ > /Users/...../Desktop/ex1.txt
--

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

I wonder if anyone can help on this one.

The parameters for LPC are:

time of the nearest frame (defaults to 0)
length of the vocal tract (defaults to .17)
use the length of the vocal tract calculated according to the
algorithm as described in Wakita(1977) (can be set to yes)

Does LPC only work on one frame?  Is that frame 'frame 0' if the default
is used?

I presume .17 is the length of a typical male vocal tract, but if you
set wakita to yes, should you change the figure or is it just ignored?

Stephen




--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 267.5.0 - Release Date: 2/06/2005

Dear Praat users,

the vocal tract lengths (VTL's) as measured from a VocalTract object are not
very reliable. The VTL numbers obtained, very much depend on the number of
formants found and the sampling frequency. Only changing the sampling
frequency and keeping the formant frequencies constant will change vocal
tract length!
A  long time ago I have implemented the Wakita VTL algorithm for
historical reasons only , and maybe I will remove it sometime in the future.
Please use VTL derived from a VocalTract with extreme care, or bettter, do not
use it at all.
I do not know of a reliable algorithm to calculate VTL from the speech signal
only. If you do please inform me.

Greetings,

David Weenink


--
David Weenink,
Institute of Phonetic Sciences, University of Amsterdam,
Herengracht 338, 1016 CG Amsterdam, The Netherlands.
phone: +31-20-5252187
email: David.Weenink@...

i write a script as followed

a=melToHertz(1000)
b=hertzToMel(1000)
echo 'a' 'b'

to my surprise, the results are 2838.3558963675514 and
569.8505624277267 not 1000 and 1000.
is there any problem?

I have 30ms segments of a fricative sound, I know how to get a Center
of gravity and central moment. However, if I want to do overlapping
windows on this segment, can the program do this automatically?
Lindsay

At 10:19 +0000 6-6-05, remilynn2000 wrote:
>a=melToHertz(1000)
>b=hertzToMel(1000)

>to my surprise, the results are 2838.3558963675514 and
>569.8505624277267 not 1000 and 1000.

there are several definitions of "mel" in the literature, it's not an
International Standard Unit or something like that. The one in Praat has an
equal number of mels and Hertzes for low frequencies. It may be useful to have
multiple definitions available in future versions of Praat.
--

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

Hello,
	 I am a relatively new Praat user interested in signal coding
in animal communication.  I do not know much about speech analysis
but so far I have used Praat to run a rudimentary analysis (source
and filter) of over 1000 bison vocalizations.   In prep for my
upcoming playback studies this summer field season, I am interested
in creating a synthesized bison bellow which would be the
mathematical average of a large number of original bison calls.  Is
there any function within Praat which allows me to create this?  So
far I only see directions indicating how to synthesize the
source/filter info from 1 to 2 (if using `convolve') sounds.
         Thanks very much for your time!
		 Megan

Hello everyone,

I have some background in C/C++ programming but not
much in compilers though yet. And I would like to
learn how to incorporate praat into C++ programs.

So recently I downloaded the source code, unpacked it
and ran "make", everything compiled fine.
I have found the tutorial but still confused on how to
create a simple script so that I could run it from
command line on Unix. For example, can I use a regular
editor (pico, etc.) to write the scripts, should they
always have .c extension? and do I compile it with all
the Makefiles, would regular gcc -c ... work?

Let's say I want to write a script that has a .wav
file as input and modifies the duration or frequency
of the sound, is it possible to do? How would I begin?
which directories do I need to look into for classes
and so on?
If I use that script in some other C++ program can I
just call it as: system("script_name"); ?

Thanks for your help in advance.

Olya.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

At 23:32 -0700 8-6-05, Olga Krachina wrote:
>I have found the tutorial but still confused on how to
>create a simple script so that I could run it from
>command line on Unix.

you use any editor (or Praat) to write the script.

>should they
>always have .c extension? and do I compile it with all
>the Makefiles, would regular gcc -c ... work?

no, Praat scripts are not C programs. You run them with Praat. From the command
line, you can do:

    praat myScript.txt

or so.

>Let's say I want to write a script that has a .wav
>file as input and modifies the duration or frequency
>of the sound, is it possible to do? How would I begin?
>which directories do I need to look into for classes
>and so on?

You call it like

    praat lengthen.praat hello.wav

where lengthen.praat contains the following text:

form Lengthen
    text fileName xx.wav
endform
Read from file... 'fileName$'
Lengthen (psola)... 1.4
Play

>If I use that script in some other C++ program can I
>just call it as: system("script_name"); ?

system ("praat lengthen.praat hello.wav");

For all those file names (including "praat") you may have to suppy the complete
paths.
--

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

I am trying to take vlt over an utterance:  If I do this:

form  xyz
             word sFile
             word name
             word prtFile
endform

# just to empty the file
filedelete 'prtFile$'

# to read the file

Read from file... 'sFile$'
select Sound 'name$'
Play


# to extract the vocal tract length
# figures are settings

select Sound 'name$'
# don't forget Nyquist - this is 8k sampling therefore 4k limit and 4 formants
# men app 1000 each women slightly less
Resample... 10000 50
To LPC (covariance)... 20 0.025 0.005 50


             To VocalTract (slice)... 10 0.17 yes
             info$ = Info
             vocalTractLength = extractNumber(info$, "Vocal tract length:")
             fileappend 'prtFile$' vocalTrackLength => 'vocalTractLength:3'
'newline$'


I get a result 0.18, I presume for the 10th frame.

However if I try to do this:

.
.
.
Resample... 10000 50
To LPC (covariance)... 20 0.025 0.005 50

n = Get number of frames
for i to n
             To VocalTract (slice)... i 0.17 yes
             info$ = Info
             vocalTractLength = extractNumber(info$, "Vocal tract length:")
             fileappend 'prtFile$' vocalTrackLength => 'vocalTractLength:3'
'newline$'
endfor

I get this:

C:\Praat>praatcon length.script 1.wav 1 out.txt
Error: Command "To VocalTract (slice)..." not available for current selection.
Script line 28 not performed or completed:
½ To VocalTract (slice)... i 0.17 yes  ╗
Script "C:\Praat\length.script" not completed.
Praatcon: command file "C:\Praat\length.script 1.wav 1 out.txt" not completed.

Can anyone tell me how to make this work through the utterance?

Stephen

--
Internal Virus Database is out-of-date.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 267.5.0 - Release Date: 2/06/2005



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

At 13:49 +1000 10-6-05, Stephen Choularton wrote:
>To LPC (covariance)... 20 0.025 0.005 50
>n = Get number of frames
>for i to n
>            To VocalTract (slice)... i 0.17 yes
>            info$ = Info
>            vocalTractLength = extractNumber(info$, "Vocal tract length:")
>            fileappend 'prtFile$' vocalTrackLength => 'vocalTractLength:3'
'newline$'
>endfor

You have to change the selection back to the LPC object:

To LPC (covariance)... 20 0.025 0.005 50
lpc = selected ("LPC")
n = Get number of frames
for i to n
             To VocalTract (slice)... i 0.17 yes
             info$ = Info
             vocalTractLength = extractNumber(info$, "Vocal tract length:")
             fileappend 'prtFile$' vocalTrackLength => 'vocalTractLength:3'
'newline$'
             Remove
             select lpc
endfor

(this also throws away all those little VocalTract objects)
--

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

Can a spectrum be converted into logHz? Also, how exactly do you use
the Bark Filter?
Lindsay

Dear all,

I have two questions:
1. I'm writing a script to insert labels in a point tier. In this
script, I want to put labels at the 25%, 50% and 75% points between
two assigned labels/points. When I ran this script (see below), it
inserted points correctly only in the last TextGrid file. In the
other files, the points seem to go pretty much anywhere. What do I
miss?

2. How do I get the name of the files in Praat objects (either Sound
or TextGrid) in printline, instead of the number of the files? I've
looked through various scripts and tried them, but I don't seem to
have found the right one.

Here is the script to insert points:
--------------------------------------
select all
numberOfSelectedTextGrids = numberOfSelected ("TextGrid")
for i to numberOfSelectedTextGrids
    textgrid'i' = selected ("TextGrid", i)
endfor

for i to numberOfSelectedTextGrids
    select textgrid'i'
    x = Get time of point... 3 2
    y = Get time of point... 3 3

    a = ('y'-'x')/4
    b = 'a'*2
    c = 'a'*3
    lab1 = 'x'+'a'
    lab2 = 'x'+'b'
    lab3 = 'x'+'c'
endfor

for i to numberOfSelectedTextGrids
    select textgrid'i'
    Insert point... 3 'lab1' BC2+25
    Insert point... 3 'lab2' BC2+50
    Insert point... 3 'lab3' BC2+75
endfor

printline inserted labels to TextGrid 'i'
---------------------------------------------
Help would be really appreciated.

Niken Adisasmito-Smith

Dear list members,

I'm collecting samples of spontaneous talk. I get many instances of
two or more people speaking at the same time. I was wondering, is
there any way of separating overlapping acoustic signals?

Thanks.

Andrea Perticone
kukypert@...

I think it will be a problem to use i for the variable in both
loops.  Let's say that you have 5 textgrids.  I think once you get down to
the second loop, the variable i will evaluate to 5 the very first time the
loop runs.  You say that the points were inserted seemingly randomly in all
but the last one -- might they all have been  inserted at the same time
indices, and they only appear to be appropriate for the last one?  I would
change to a different letter for the second loop.

To get the name of an object, instead of its number, you can do something like:

select TextGrid'i'
object_name$ = selected$ ("TextGrid")
printline inserted labels to 'object_name$'

Is that what you wanted to do?

Best of luck,

Katherine.

At 22:22 +0000 10-6-05, adisasmito_smith wrote:
>select all
>numberOfSelectedTextGrids = numberOfSelected ("TextGrid")
>for i to numberOfSelectedTextGrids
>   textgrid'i' = selected ("TextGrid", i)
>endfor

This is correct.

>for i to numberOfSelectedTextGrids
>   select textgrid'i'
>   x = Get time of point... 3 2
>   y = Get time of point... 3 3
>
>   a = ('y'-'x')/4
>   b = 'a'*2
>   c = 'a'*3
>   lab1 = 'x'+'a'
>   lab2 = 'x'+'b'
>   lab3 = 'x'+'c'
>endfor

But this is not. lab1 will contain a time in the last TextGrid.

>for i to numberOfSelectedTextGrids
>   select textgrid'i'
>   Insert point... 3 'lab1' BC2+25
>   Insert point... 3 'lab2' BC2+50
>   Insert point... 3 'lab3' BC2+75
>endfor

... so that all of these points will be inserted at the same time in all the
TextGrids! A general change would be:

>   lab1'i' = x + a
>   lab2'i' = x + b
>   lab3'i' = x + c
...
>   Insert point... 3 lab1'i' BC2+25
>   Insert point... 3 lab2'i' BC2+50
>   Insert point... 3 lab3'i' BC2+75

although the following would also simply work:

>for i to numberOfSelectedTextGrids
>   select textgrid'i'
>   x = Get time of point... 3 2
>   y = Get time of point... 3 3
>
>   a = (y - x)/4
>   b = a * 2
>   c = a * 3
>   lab1 = x + a
>   lab2 = x + b
>   lab3 = x + c
>   Insert point... 3 lab1 BC2+25
>   Insert point... 3 lab2 BC2+50
>   Insert point... 3 lab3 BC2+75
>endfor

(Please note that there is no problem in reusing the variable i. It will start
at 1 each time.)
--

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

dear praat users


i am doing my final msc thesis and my thesis title is that " vocal
tract ailments diagonsis trough voice feature analysis".
the idea is that here i want to compare both voice samples of the
same person(sick and healthy),the problem is that i faced in these
days , which acoustic features of voice i sould consider,bcoz lot of
ambiguties are exist in the features which i selected  like
jitter,shimmer ,frequency,picth bandwidth,intensity and lot of others.

plz help me

zafar iqbal

I would like to experiment with real time analysis.  So as a first
step, I am trying to recompile with MINGW compiler.  Not having any
luck.  Does anybody have an example/tutorial how to overcome the
missing headers, etc?

Thanks.
David

Hello all,

I am writing a Praat resynthesis script, but I'm having two problems:

(1) How can I decrease intensity?
I'm creating an IntensityTier, adding points with positive or negative
values, and using "Multiply" with a Sound. The result is always an
increase.

(2) How can I modify the time span of an interval in a TextGrid?
Assuming I change the duration of a particular word inside an
utterance Sound, and then resynthesize the utterance, I would like to
modify the interval boundaries of the word in the TextGrid to reflect
the modified duration in the Sound. (The manual equivalent would be
selecting and dragging the boundary in the TextGrid.)

Thank you,
Cecilia

Hi



Could anybody help me?



What I am currently trying to do is to create a vowel continuum through varying
the F1 value in a range that goes from 300 to 600 Hz.

I have tried the Praat source-filter synthesis, thus creating a formant tier and
associating it to a source which is either extracted from a sound or created
from scratch.

The problem I am encountering is that when I create a sound associating the
formant tier to the source signal and query its F1 value (or the value of  any
one of the formants),  I am getting  a different value  from the one I had
previously fixed in the formant tier.

I am a novice and I can not imagine an explanation for this discrepancy other
than some kind of effect the source could have over the filter.

What could be wrong? Would you have any suggestions of what could have been done
differently?

How is it possible to create a sound with the “exact” formant value I intend?





---------------------------------
Yahoo! Mail: agora com 1GB de espaço grátis. Abra sua conta!

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

hi could anyone just mail me a copy of hermes "spectral harmonic summation
method for pitch detection" (the paper.)? or a link would do.
Thanks,
ANI.


---------------------------------
Discover Yahoo!
  Get on-the-go sports scores, stock quotes, news & more. Check it out!

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

At 23:58 +0000 14-6-05, ovesdotter1 wrote:
>(1) How can I decrease intensity?
>I'm creating an IntensityTier, adding points with positive or negative
>values, and using "Multiply" with a Sound. The result is always an
>increase.

in Praat 4.3.14, you can switch off automatic scaling in the Multiply window. An
IntensityTier with a value of -6 dB will then halve the intensity.

>(2) How can I modify the time span of an interval in a TextGrid?
>Assuming I change the duration of a particular word inside an
>utterance Sound, and then resynthesize the utterance, I would like to
>modify the interval boundaries of the word in the TextGrid to reflect
>the modified duration in the Sound. (The manual equivalent would be
>selecting and dragging the boundary in the TextGrid.)

I see. Praat cannot do that yet. A future solution would be to that you can
select a DurationTier and a TextGrid together, then see a button "Change times"
or so. Please keep asking me about this, in case I forget.

best wishes,
Paul
--

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 21:44 -0300 14-6-05, Daniel Silva wrote:
>I have tried the Praat source-filter synthesis, thus creating a formant tier
and associating it to a source which is either extracted from a sound or created
from scratch.
>
>The problem I am encountering is that when I create a sound associating the
formant tier to the source signal and query its F1 value (or the value of  any
one of the formants),  I am getting  a different value  from the one I had
previously fixed in the formant tier.

The formant analysis assumes that the envelope of the source signal has a slope
of -6 dB per octave.

>How is it possible to create a sound with the “exact” formant value I intend?

you should get close to that (though never exactly) if the envelop of your
source sound has the required spectral slope. Try to create a pulse train, then
filter that in the frequency domain (there's a button for that) with the formula
"self/(x+50)".

Please also make sure that you filter your source signal with many more formants
than you want to measure, i.e. at least give it an F6, F7, F8 and F9 with
reasonable values.
--

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