2.2.3 Analysing
By looking at the top menu, you can now choose which specific feature of your sound file to look at. To begin with, simply skip to the first word of your sound file by marking it in the waveform window and clicking on the “sel” button in the lower left menu. This button automatically zooms in on the part you have marked. Now you can look closely at the sounds which are contained within the spoken word. Mark them and begin inserting boundaries as described in Section 2.2.2. In order to tell consonants and vowels apart, remember that the former consist mainly of aperiodic waves while the latter contain periodic waves.
Praat enables you to look closely at four features of the sound file:
For a basic acoustic analysis you will especially need the pitch, intensity and formant functions.
To start with, we can find clues about the speaker by just looking at the wave form of the sound data. We can for example measure the length of vowels or syllables to draw conclusions about the speech rate. Also, the voice onset time can be measured by reading the wave form.
The pitch function in Praat refers to the fundamental frequency of the sound file. By looking at the pitch we can distinguish properties of intonation and word stress. By looking at the maxima and minima of the pitch in a sentence we can measure the so called small and wide pitch range. To get the average wide pitch range we need to measure the highest initial pitch and the lowest final pitch. Then we need to subtract the final low value from the initial high value to get the wide pitch range. In order to get the small pitch range, we need to do the same as before but now with non-initial highs and non-final lows .
The intensity too can tell us something about word stress within a sentence since the stressed syllables have a higher intensity than the unstressed syllables.
To draw conclusions about the quality of vowels we need to look at the formant frequencies. By clicking on “Show Formants” under “Formants” in the top menu, Praat shows the formants in the spectrum as little red dots. To get the actual frequency of a formant we need to mark the place where we want to measure the formant and just press the F1 key for the F1 value, the F2 key for the F2 value and so on. Make sure that you are measuring the vowel formants right in the middle of the vowel to get a clear value which is not affected by transition.
When you have sufficient values of various vowels you can start to plot them. With Microsoft Excel or Open Office Calc you can create a diagram which contains the F1 and F2 values you found. To get a diagram which resembles the classical vowel chart you need to change the axis so the x value rises from right to left and the y value from top to bottom. You can get good speaker specific results by just looking at the F1 and F2 values, but if you are comparing two speakers with each other you will need to normalize the values in order to get decent results. Since everyone has different physiological preconditions (i.e. shape of head, neck and oral/nasal cavities) this will lead to slightly different realizations of the same vowel. There are various methods for the vowel normalization. For detailed information on these methods visit http://ncslaap.lib.ncsu.edu/tools/norm/index.php. There, you can upload the formant values found and get conclusive charts of your data. We already know that in English the first two formants are sufficient to get results, but since the “Norm” website is designed for linguists of different languages you need to provide the F3 values, as well.
After the first few times working with Praat you will recognise that the basic steps are easy to recreate. Praat is an easy-to-handle way to do your phonetic research at home and by yourself. It requires a huge amount of time to gather all the data from a sound file and there are many mistakes you can make during your analysis. So one has to work very patiently and accurately in order to get reliable results.
Created with the Personal Edition of HelpNDoc: Generate EPub eBooks with ease