Aperiodicity cue for fricative voicing contrasts

Area of Science:

• Phonetics and Acoustics
• Speech Science
• Computational Linguistics

Background:

• Voicing contrasts in fricatives are crucial for speech intelligibility.
• Traditional acoustic features for voicing analysis have limitations.
• French fricative phoneme characterization requires robust acoustic cues.

Purpose of the Study:

• To introduce and evaluate a novel aperiodicity metric for classifying French fricatives.
• To assess the metric's effectiveness in distinguishing voiced and unvoiced fricatives.
• To determine if aperiodicity enhances existing acoustic features for phoneme classification.

Main Methods:

• Development of an aperiodicity metric from the YIN difference function.
• Analysis of controlled, noise-free French consonant-vowel-consonant-vowel (CVCV) speech.
• Classification using the aperiodicity metric, root mean square amplitude, and standard classifiers.

Main Results:

• The aperiodicity metric reliably and objectively classifies fricatives as voiced or unvoiced.
• Combining aperiodicity with root mean square amplitude achieved approximately 83% classification accuracy.
• The metric provided complementary information to conventional acoustic features.

Conclusions:

• Aperiodicity-based acoustic cues are valuable for detailed phoneme characterization.
• The proposed metric offers a reliable method for analyzing voicing in French fricatives.
• This approach enhances the understanding of acoustic properties in speech signals.