contains WAV
speech data file , several KB long
The analysis of the electroglottographic signal was done using the description of the waveform proposed in section 12. In order to attain a direct comparison with other methods for measuring laryngeal behavior, the same data as in Claßen et al. (1996) was used. In the further analysis not only stress and tenseness factors were investigated, but also different vowel types as well as speakers' gender. This was done in order to get a wider perspective of the proposed parametrization and to provide a broader basis for the comparison with other methods of measuring laryngeal activity.
The following near-minimal pairs involving tense and lax vowels were selected. The segmental context [th_l] was used throughout the experiment (Table 9).
For each of the ten words, the variants with either stressed or unstressed target vowels were stimulated by appending the suffixes -isch and -ist, respectively (eg. Ventil ~ ventilisch, ventilist). This was used to introduce the following combinations of tenseness and stress:
* tense stressed (anatolisch, klientelisch, thulisch, ventilisch, spitalisch)
* lax stressed (kartellisch, metallisch, tormentillisch, schatullisch, ayatollisch)
* tense unstressed (spitalist, ventilist, anatolist, klientelist, thulist)
* lax unstressed (kartellist, metallist, ayatollist, tormentillist, schatullist)
Note: contains WAV
speech data file , several KB long
| Table 9: Words used in the experiment | ||
| vowel group | tense vowel | lax vowel |
| /a/ | Spital (hospital) [a]
|
Metall (metal) [a]
|
| /e/ | Klientel (client) [e]
|
Kartell (cartell) []
|
| /i/ | Ventil (vent) [i]
|
Tormentill (tormentilla) [I]
|
| /o/ | Anatolien (Anatolia) [o]
|
Ayatollah (ayatollah) []
|
| /u/ | Thulium (thulium) [u]
|
Schatulle (casket) []
|
In comparison to Sluijter's (1995) experiments it should be taken into consideration that in her investigation of word stress vowels were placed in both focus and non-focus positions. The tokens examined in this study are restricted to focal positions ([+F]).
Each of the resulting 20 words was read twice by ten native speakers of German, five female and five male, all with no history of speech, voice or hearing disorders. Their ages ranged from 24 to 34 and none of them smoked. 400 tokens were available.
Recordings were made in two channels: the first channel contained the speech signal and the second the EGG signal. No special acoustic precautions were taken in the recording room. The signals were sampled at 16 kHz with 16-bit resolution and stored directly on a computer disk to avoid any signal distortion. The acoustic signals were recorded with a Sony ECM 428 condenser microphone connected to an A/D converter with an built-in preamplifier (Ariel Proport Model 656). The recording program was integrated in the S_Tools V5 speech processing environment (Deutsch & Noll, 1993).
The EGG signal was supplied by a Laryngograph processor (from Laryngograph Ltd.). Neither additional filtering (except for low pass filtering for sampling, fl=8 kHz)) nor signal modification were performed.
The amplification of the EGG signal, as well as of the acoustic signal, was adjusted to the subjects' speech. The acoustic and EGG recordings were not calibrated.
A recording of one female speaker was rejected due to poor quality (the signal was notably weaker than in the other recordings, perhaps due to incorrect positioning of the electrodes).
