Optimization-based modeling of Lombard speech articulation: Supraglottal characteristics

Area of Science:

• Speech Production
• Acoustic Phonetics
• Psychoacoustics

Background:

• The Lombard effect describes speech modifications in noise.
• Previous models have not fully explained the articulatory basis of these changes.

Purpose of the Study:

• To investigate a simplified model of speech production.
• To determine if articulatory effort optimization can predict Lombard speech features.
• To test the listener-oriented hypothesis of speech adaptation in noise.

Main Methods:

• Simulations of static vowels under varying background noise levels.
• A model optimizing articulatory effort against intelligibility.
• Analysis of predicted articulatory and acoustic modifications.

Main Results:

• The model successfully predicted articulatory and acoustic changes observed in Lombard speech.
• These predictions were achieved when articulatory effort constraints decreased with increasing noise.
• The model highlights the trade-off between effort and intelligibility.

Conclusions:

• A simplified optimization model can account for Lombard speech phenomena.
• Lombard speech appears to be listener-oriented, with speakers adapting articulation for clarity in noise.
• This suggests speakers actively modify speech production based on environmental acoustic conditions.