Linguistics 520: Lab Assignment #5

9/25/2013: Due 10/2/2013

Goals:

1. Use formant analysis to quantify the quality of American English vowels.
2. Characterize individual and group variation in vowel formants.
3. Learn an appropriate method for normalizing vowel formant measurements.

Initial Background:

Read (and understand) the Wikipedia article about John Wells' "lexical sets".

Recording:

(If you're a native speaker of American English, record these yourself. If you aren't, recruit a friend or acquaintance who is. The recording should only take a couple of minutes.)

Choose (and download) one of 10 randomizations of 12 /h.VOWEL.d/ monosyllabic English words, in the carrier phrase "Now say __ again: 1 2 3 4 5 6 7 8 9 10.

For example, Vowels3.txt is

     (01) Now say "heed" again.		[rhymes with "need"; FLEECE lexical set]
     (02) Now say "hayed" again.	[rhymes with "made"; FACE lexical set]
     (03) Now say "who'd" again. 	[rhymes with "mood"; GOOSE lexical set]
     (04) Now say "had" again.		[rhymes with "bad"; TRAP lexical set]
     (05) Now say "head" again.		[rhymes with "bed"; DRESS lexical set]
     (06) Now say "hod" again.		[rhymes with "nod"; LOT lexical set]
     (07) Now say "hud" again. 		[rhymes with "bud"; STRUT lexical set]
     (08) Now say "hoed" again.		[rhymes with "load"; GOAT lexical set]
     (09) Now say "hawed" again.	[rhymes with "Maud"; THOUGHT lexical set]
     (10) Now say "hid" again.		[rhymes with "bid"; KIT lexical set]
     (11) Now say "heard" again.	[rhymes with "bird"; NURSE lexical set]
     (12) Now say "hood" again.		[rhymes with "good"; FOOT lexical set]

Print out your chosen randomization, using a large enough font for it to be easy to read; or arrange to display it in a convenient and easy-to-read browser window. Study the items in the list briefly, to be sure that you know how to say them.

In a quiet environment, record the twelve sentences twice. I suggest a format like:

Number one. Now say 'heed' again.
Number two. Now say 'hayed' again.
etc.

Don't read the "rhymes with..." part -- that's just to clarify what the sound is supposed to be. Also, many of you (or your chosen talkers) will merge the THOUGHT and LOT lexical sets. If so, don't worry about it.

Record single-channel .wav files at 22050 Hz sampling rate. Don't record in stereo; don't record into an .mp3 file. If you don't know how to do this, ask.

Measurement:

Create 24 files, one for each of the two readings of the 12 vowels. I suggest naming the files for the lexical sets involved, e.g. the two repetitions of "Now say hood again" would go into FOOT1.wav and FOOT2.wav.

Now use Praat to make the following measurements:

1. Duration of the vowel:

2. Average f0 of the vowel (via Pitch>>Get Pitch)

3. F1, F2, F3 at 20%, 50%, 80% of the vowel duration.

(Place the cursor at the appropriate time, and use Formants>>Get Formant Listing)

Copy the duration, f0 and formant values into a file, keeping track of which file you used and where you made the measurements -- e.g.

file      start  end dur f0  t.20  F1.20  F2.20  F3.20 t.50 F1.50 F2.50 F3.50 t.80 F1.80 F2.80 F3.80
FACE1.wav 792 980 188 117 830 440 2008 2584 886 391 1998 2598 942 361 2050 2573

If you were doing this on a large scale, it would make sense to set up a Praat script to lead you through the process more conveniently -- or even to set up a system for automating the formant tracking entirely. We'll see some examples of such techniques later in the course. But for this lab, it will do you good to do everything by hand...

Post-Recording Background -- Read two classic papers:

Gordon Peterson and Harold Barney, "Control Methods Used in a Study of the Vowels", JASA 1952.

James Hillenbrand, Laura A. Getty, Michael J. Clark, and Kimberlee Wheeler, "Acoustic characteristics of American English vowels", JASA 1995.

Additional data for use in the next lab

You'll compare your formant space with the measurements in the Peterson-Barney and Hillenbrand-et-al. papers.

The formant measurements from Peterson & Barney 1952 are in the file pbtable, which you should download to a folder accessible to R.

The table columns are:

1: m = man, w=woman, c=child
2: m=male, f=female
3: N=speaker number
4: V=vowel name (heed/iy, hid/ih, head/eh, had/ae, hod/aa, hawed/ao, hood/uh, who'd/uw, hud/ah, heard/er)
5: F0 in Hz
6: F1
7: F2
8: F3


The formant measurements from Hillenbrand et al. 1995 are in a file htable, which you should download to a folder accessible to R.

The table columns are:

1: m=man, w=woman, b=boy, g=girl
2: talker number
3: vowel (ae="had", ah="hod", aw="hawed", eh="head", er="heard", ey="haid" (="hayed"), ih="hid", iy="heed", oa=/o/ as in "boat", oo="hood", uh="hud", uw="who'd")
4: duration in msec
5: f0 at "steady state"
6: F1 at "steady state"
7: F2 at "steady state"
8: F3 at "steady state"
9: F4 at "steady state"
10: F1 at 20% of vowel duration
11: F2 at 20% of vowel duration
12: F3 at 20% of vowel duration
13: F1 at 50% of vowel duration
14: F2 at 50% of vowel duration
15: F3 at 50% of vowel duration
16: F1 at 80% of vowel duration
17: F2 at 80% of vowel duration
18: F3 at 80% of vowel duration

(In this table, "0" as a formant measurement means that no measurement was available.)

Note that the vowel labels in the two datasets are inconsistent :-(. We'll straighten this out later.