A quick guide to PRAAT Yap Ngee Thai 8 November 2011
11-12-15
© Ngee Thai, Yap
1
What is PRAAT? • A computer software for acoustic analysis • Developed by Paul Boersma and David Weenink at the University of Amsterdam
• Available for free from http://www.fon.hum.uva.nl/praat/
11-12-15
© Ngee Thai, Yap
2
What can you use PRAAT for? • • • •
To record speech sounds To play back and edit speech sounds To view waveforms and spectrogram To extract acoustic information about specific speech sounds • To construct speech perception experiments (I teach this to postgrad students but if you are really interested, you can do this for undergrad project paper too). 11-12-15
© Ngee Thai, Yap
3
Example: Waveforms and Spectrogram for sit
11-12-15
[s]
[ Ngee ] Thai, Yap ©
I
[th]
4
What acoustic information can you get from PRAAT? • • • • •
Closure duration for stops Vowel/consonant/word duration Pitch – fundamental frequency (F0) Intensity Formant values for vowels e.g.First formant (F1) High vowels have lower F1 values; low vowels have higher F1 values
11-12-15
© Ngee Thai, Yap
5
How can you use the acoustic information obtained? • To objectively show difference between varieties spoken by two groups of speakers • To objectively show individual differences in speech when speaking to different interlocutors. – E.g. Pitch differences speaking/reading to children vs. adults speaking to your boss vs. your boy/girlfriend 11-12-15
© Ngee Thai, Yap
6
Summary of Tutorial Topics 1. Downloading and installing PRAAT 2. Recording sound stimuli with PRAAT 3. Viewing and extracting acoustic information from sound files 4. Selecting, splicing and saving selections from a sound file
11-12-15
© Ngee Thai, Yap
7
1. Downloading and installing PRAAT •
Check your computer Operating System (OS) – Windows – Apple (macintosh) – Linux etc.
•
Go to the relevant download page from the PRAAT’s main page http://www.fon.hum.uva.nl/praat/
•
If you use windows, choose a suitable bit version for your computer or laptop (most computers can run the 32-bit version) unless your computer uses an OS older than Window 2000, then scroll to the bottom of the page for an older version of PRAAT
•
Save the download file in a location that you can remember, or save it on your desktop. The download file is a zipped folder. You will need to extract the files onto a different folder. You can either create a folder on desktop or in your program folder. The location doesn’t really matter. You just need to remember where you have saved it.
11-12-15
© Ngee Thai, Yap
8
PRAAT ICON • If you are successful in downloading PRAAT, you will see the following icon either on your desktop or in the folder where you have saved PRAAT. Double click on the icon to start using PRAAT.
11-12-15
© Ngee Thai, Yap
9
Checking out PRAAT • Double click the praat icon. • Two windows will open up: Praat object and Praat picture. • You can close praat picture. We use this for drawing only. • From the praat object window, check out the pull down menus. – – – – 11-12-15
Praat menu New menu Read menu Write menu
(for constructing experiments) (for recording speech sounds) (to open and view speech sounds) (to save recordings or work done) © Ngee Thai, Yap
10
2a. Recording sound stimuli with PRAAT • Before you start make sure you have a good line-in microphone. A very good noisecancelling microphone can cost about RM150300 • Make sure you have disabled your default microphone from your laptop or it will interfere with the recording.
11-12-15
© Ngee Thai, Yap
11
2b. Recording sound stimuli with PRAAT • To start recording with PRAAT, from the new menu, click record mono sound. • A sound recorder window will pop up. • Select sampling frequency between 22,050 Hz to 44,100 Hz This determines the quality of the recordings and the type of acoustic analysis you can do with it. But the higher you go the bigger the file (in kbs). • Click record and start speaking. If the microphone is working correctly, you will see shades of green, yellow and red on the meter box. When done, click stop, name the file and click save to list. • On the praat object window, a new object will appear. You may now click play to listen to the recording, or click edit to inspect the waveform and the spectrogram.
11-12-15
© Ngee Thai, Yap
12
Sound recorder window
11-12-15
© Ngee Thai, Yap
13
3a.
Viewing and extracting acoustic information from sound files
•
It’s important to choose the right words, sentences to record. What you would record depends on the type of information you want to collect.
•
E.g. If you want to show the way Malaysians pronounces voiceless stops (usually without aspiration) you may record words with voiceless stops in various positions E.g.
Initial medial final
/pɪt/ /hæpɪ/ /kæp/
‘pit’ ‘happy’ ‘cap’
Notice I am controlling for the type of vowels before and after the /p/. To know when we start and end the /p/, we may need to record these words in a carrier phrase, for eg. “Please say ________ again” /pli:z seɪ ________ ǝgeɪn / Which acoustic measure will benefit from recording the stimuli with a carrier phrase?
11-12-15
© Ngee Thai, Yap
14
How to check recording quality? Spot the difference.
11-12-15
© Ngee Thai, Yap
15
Applying your knowledge in phonetics What information can you get from this recording of ‘tip’?
Closure duration of /p/
VOT
11-12-15
Vowel duration
© Ngee Thai, Yap
Aspirated stop in word final position
16
3b. Viewing and extracting acoustic information from sound files • Familiarize yourself with the menus in the editing window Identify these. What information do they show? • Blue vertical lines in the waveform window • Blue connected dot in the spectrogram • Yellow line in the spectrogram • Red dots in the spectrogram that seem to form four distinct lines 11-12-15
© Ngee Thai, Yap
17
3b. Viewing and extracting acoustic information from sound files • Blue vertical lines in the waveform window Pulses • Blue connected dot in the spectrogram Pitch • Yellow line in the spectrogram Intensity • Red dots in the spectrogram that seem to form four distinct lines Vowel formants 11-12-15
© Ngee Thai, Yap
18
What specific measure can you get from PRAAT and how? • Durations – E.g.
stop durations vowel durations voice onset time
11-12-15
© Ngee Thai, Yap
19
4a. Selecting, splicing and saving selections from a sound file • When you record the soundfiles, you may record everything at one time. You may not want to interfere and record sentence by sentence, or word by word. • So after the recording session, you will need to cut out what you need and save them for further work to be done with it. 11-12-15
© Ngee Thai, Yap
20
4b. Selecting, splicing and saving selections from a sound file • This is done in the editing window. • Using the mouse, click and drag to select the target word from the recorded sentence. • For example if you recorded “Say ____ again” and the target words is pit • You should select the section from the end of the vowel in say to the just before the schwa begins in again. 11-12-15
© Ngee Thai, Yap
21
All three sentences
11-12-15
© Ngee Thai, Yap
22
Say pit again Is this selection correct? Explain based on your understanding in phonetics
pit
11-12-15
© Ngee Thai, Yap
23
4c. Selecting, splicing and saving selections from a sound file • With the selection, go to the file menu and click extract selected sound (time from 0) • A new object will appear in the object window. You can rename this object as pit. • To save this sound file, click write from the praat object and select write as wav file. Be sure to remember the folder where this file is saved. NOTE:
11-12-15
When you close the praat object window, everything is lost unless you have done the write function. © Ngee Thai, Yap
24
Some acoustic measures for /p/ in pit Closure duration 181 ms
VOT 104ms
11-12-15
© Ngee Thai, Yap
25
Ready for some hands-on practice? • Record the sentences • Splice out the required word • Compare the acoustic values for closure duration and VOT for the /p/ in the three different environments. • Compare your results with your friends, those with similar L1 background and those with different L1 background 11-12-15
© Yap, Ngee Thai
26
Want to do more serious research? • Collect the results from your classmates • Or get speakers with two different backgrounds to record these sentences e.g. Persian ESL speakers Malay ESL speakers Mandarin ESL speakers • Collect some background information of the speakers – Mother tongue and dominant language – English language proficiency
• Pool the results by averaging the acoustic values • Present the results in a chart/boxplot (use excel/spss) • Run a simple statistical test t-test to see if difference between group is significant 11-12-15
© Ngee Thai, Yap
27