Perceptual normalization of vowels produced by sinusoidal voices

Area of Science:

• Acoustic Phonetics
• Auditory Perception
• Speech Science

Background:

• Listeners perceive linguistic properties in sinusoidal replicas of sentences, despite lacking typical vocal acoustic cues.
• The study investigates whether this perception relies on a novel strategy or standard speech processing mechanisms.

Purpose of the Study:

• To determine if sinusoidal speech signals are processed via ordinary speech perception mechanisms.
• To test if perception of sinusoidal signals is sensitive to vocal-tract dimensions.

Main Methods:

• Participants identified target /bVt/ syllables in sentences composed of sinusoidal tones.
• Sentences varied in the frequency range of component tones, an acoustic correlate of vocal-tract scale.
• Vowel quality perception was assessed.

Main Results:

• Vowel quality perception was influenced by the frequency variation, indicating vocal-tract normalization.
• Perception of sinusoidal vowels depended on the relationship among tones, not individual tone phonetic similarity.
• Findings support that sinusoidal replicas are phonetically perceptible due to preserved time-varying information.

Conclusions:

• Sinusoidal speech replicas are perceived using standard speech processing, including vocal-tract normalization.
• The perception relies on the relational information within the signal, not isolated phonetic qualities.
• Preservation of time-varying acoustic information is key to the phonetic intelligibility of sinusoidal speech.