EUSTACE
HOME |
DESIGN |
EXAMPLES |
SENTENCES |
PROCEDURE |
RESULTS |
DISSERTATION |
LICENCE |
DOWNLOAD |
GLOSSARY |
|
---|
Procedure: Organisation of recorded sentences into sound files
Utterances were extracted from the original recordings and merged into sound files with other sentences for the same speaker and keyword, to facilitate the application of consistent measurement criteria for each test syllable. There are four sound files for each keyword for each subject:
- 2 repetitions X 7 sentences, utterance-medial, unaccented.
- 2 repetitions X 7 sentences, utterance-medial, accented.
- 2 repetitions X 5 sentences, utterance-edge, accented.
- 2 repetitions X 5 sentences, utterance-edge, accented.
The names of the sound files provide information about:
- the speaker.
- the type of keyword, right-headed or left-headed.
- the keyword identity, this label being the first three letters of the keyword name, as shown in the experimental design (see DESIGN page).
- the position of the keyword, utterance-medial or utterance-edge.
- the accent condition of the keyword, accented or unaccented.
Table 1 illustrates these conventions, which are also used for the
labels files.
Interpretation | Range of values |
---|---|
Speaker identity | f1 f2 f3 m1 m2 m3 |
Keyword type | l r |
Keyword identity | cap dog fis mac par
sen spe ten com dis jui mai men por sen sup |
Utterance position | e m |
Accent condition | a u |
Filenames are derived by running these codes together: for example,
``f1rsenea.wav'' is a speech file containing female speaker F1's utterances
for the right-headed keyword send in the utterance-edge
condition with an accented keyword. A full list of the available
files, with the interpretation of each file name, is available (see
FULL LIST page).
Up to PROCEDURE page | Forward to FULL LIST page |