CS4706 - Spring 2008
Homework 1 - "Using Praat"


Due: Wednesday, February 13, 2008, by 2:40pm.
Submit in Courseworks

Solve all the exercises using Praat, which can be downloaded from http://www.praat.org.
You can use the following documentation:

(1) Analysis of sound files (20 points)

The files win1.wav and win2.wav have the same words, "Anna will win Manny", but uttered differently. Describe the acoustic differences that might account for the different meanings (e.g., variations in pitch and intensity). Provide evidence to support your claims. (Approx. three or four paragraphs.)

(2) Manipulation of sound files (40 points)

The file thermometer.wav contains a declarative contour. Modify its pitch and duration to produce two contours that convey different meanings (yes/no question, incredulity, stress on a different word, etc. —use your imagination!).

Submit the resulting .wav files.

Hint:
a) Open the file thermometer.wav.
b) Create a Manipulation object, and Edit it.
c) Select Pitch → Stylize Pitch (2 st).
d) Have fun modifying the pitch and the duration tiers!

(3) Phonetical Segmentation (40 points)

a) Record the phrases "It's hard to recognize speech" and "It's hard to wreck a nice beach" with your own voice. Format: mono and 16KHz sampling frequency
b) Remove any leading and trailing silence.
c) Create a TextGrid object for each file, with only one interval tier called "phones".
d) In the new tier, create and label one interval for each phone. Use the ARPAbet symbols (Jurafsky & Martin, Section 4.1).
e) Compare the two utterances and their phonetical segmentation. Do you find any differences that could be used to distinguish one from the other?
f) Learning to pronounce a new phoneme: The following four files are recordings of four Arabic words that contain the phoneme: Χ (IPA symbol), [x] (IRPAbet symbol). You are asked to record these words with your own voice. You should also submit the phonatic transcription of each word, using IRPAbet symbols. Audio file format must be mono and 16KHz sampling frequency.
1. khubz.wav (bread)
2. khabaza.wav (baked)
3. makhbadz.wav (bakery)
4. dakhala.wav (enter)

Submit the .wav and .TextGrid files, and a small report to answer point e, and the four phonatic transcriptions of the words in f) (approx. two or three paragraphs).

(4) Extra credit (10 points)

You may hand in this problem (and only this one!) either in hard copy or in electronic form in the lecture of the due date.

The following waveform and spectrogram correspond to a short phrase ellicited fluently and in a neutral tone by a male native speaker of American English.

HW1 - Problem 4

a) Identify the beginning and end of as many phones as you can, specifying whether it is a vowel or a consonant.
b) Which of the following phrases does these waveform and spectrogram correspond to?
    1. "Harry Potter can't fly."
    2. "Columbia University."
    3. "Anna likes New York."
    4. "Where did Tweety go?"
    5. "Amelia made the jam."

Hint: Record yourself saying these phrases, and compare their waveforms and spectrograms with the one shown above.