It should be noted that absolute values are not necessarily representative of real word results since the Lombard effect was not considered before mixing the speech with the noise.
As with the first study, absolute values are not necessarily representative of real word results since the Lombard effect was not considered before mixing the speech with the noise.
The Lombard effect
is probably familiar to anyone who has ever heard someone who is wearing headphones try to talk, but not all animals show it--frogs don't seem to, for instance.
The Lombard effect has been reported to cause measureable differences in vowel intensity and duration, and also in formant frequencies: ambient noise elevates the speech amplitude by 5-10 dB, increases word durations by 10-20%, and increases significantly the F1 and F2 frequencies, thus causing a shift in the vowel space (van Summers, Pisoni, Bernacki, Pedlow, Stokes 1988; Castellanos, Benedi, Casacuberta 1996; Beckford Wassink, Wright, Franklin 2007).
Second, the Lombard effect induced by the two different noise masks caused the duration of the short vowels, but not the long ones, to increase significantly.
Third, the Lombard effect resulted in an increase in the F1 of the mid-high vowels, but had no effect on the Euclidean distances of the short and long vowels.