These are notes for classes on 9/11 – 9/13. They are meant for you to be able to replicate what we did in class on your own computer, and to do further work on your own (follow links for more reading, try other things in Praat…). Some of it is based on pp. 17-27 of Erik Thomas’ excellent Sociophonetics: An Introduction (2011, Palgrave Macmillan).
An acoustic waveform display plots sound pressure (vertical axis), which is related to the amplitude or intensity of a sound, against time (horizontal axis). Here is a useful tutorial on understanding waveforms (the site also has lots of other useful tutorial information on acoustic phonetics and Praat).
Before we turn to speech, we’ll look at simpler sine waves. We can create these in Praat with New > Sound > Create Sound as pure tone. We’ll create three tones. For one, change Tone frequency to 200 hz, and leave amplitude with its default setting of 0.2 Pa, then press OK. For the second, leave Tone frequency at 200 hz, and change amplitude to 0.1 Pa. For the third, use 100 hz and 0.2 Pa.
We should now have three Sound tone objects. We want to look at them all at once, so we’ll combine them. Select them all (hold down shift on a mac), and use Combine > Concatenate. Now we’ll look at the new one by selecting it and choosing View and Edit. Try selecting portions of it and playing them.
What do you notice about the relationship between each of frequency and amplitude to the waveform display?
Pitch is measured in hertz, which is the number of cycles per second. It’s easy to see the repeating cycles (AKA periods) in a sine wave, and we can measure the frequency of these sine waves by hand. Click in a section of the waveform and select a section roughly 0.05 sec. – 1/20 of a sec. – in length (it can really be any length). Now do “Select > Move start of selection to nearest zero crossing”, and then adjust the end of the selection to be exactly 0.05 sec. after the start (Select > Select…, and change the value of the end of selection by copying from beginning of selection and adding 0.05). Now choose View > Zoom to selection. Our window should be labeled 0.05 sec. in length. If we chose a selection from one of the 200 hz sine waves, we should be able to count 10 cycles, and if we chose a selection from the 100 hz sine wave, we should be able to count 5. Alternatively, we could measure the length of a cycle in ms., and divide this into 1000. We should find that our 200 hz sine wave has periods of length 5 ms (1000/5 = 200).
Speech is much more complex than a sine wave/pure tone, but much of it does have pitch, and we can also measure its pitch in the same way as we just did. The main pitch, or fundamental frequency, of speech comes from the speed of vocal fold vibration. Here‘s an excellent example of vocal fold vibration at different pitches. Vocal fold vibration causes air molecules do compress and rarify, which is then picked up by a membrane on the ear drum, or by a microphone. In the case of the microphone, this is then transformed into an electrical waveform. In Praat, when we look at a recording, we are seeing a visualization of that electrical waveform (which has been sampled to bring it into the digital domain).
Voiced speech is not quite periodic, since the vocal folds do not vibrate at a constant rate, but it’s close enough that it usually referred to as complex periodic signal. Let’s try singing at two different pitches, and finding the frequency using the waveform. We need to look for the large repeating patterns, which are sometimes termed the glottal pulses. If we sing two notes an octave apart, we should find that the pitch doubles between them. Here‘s our classroom sample of a woman singing an octave, and here is a man. (To get an accurate automatic pitch track of the female voice, go to Pitch > Pitch Settings and increase the range to a 600 hz maximum – thanks to John Kingston for this tip).
To record, go to New > Record mono sound and hit record, then stop when you are done. You need to adjust the level of your recording outside of Praat. Once you have a recording you want to look at and/or save, choose Save to list. You’ll then have a new Sound object in your Object list.
Speech also contains aperiodic signals (non-periodic), and these are of two types. A sustained aperiodic signal is called noise. Fricatives produce noise. A non-sustained aperiodic signal is a transient. The release, or burst, of stops, produces a transient. Let’s record initially stressed [?p?], [?b?], [?m?], [?s?],[?f?], [?v?], and [?w?] and look at the differences amongst them.