Formant-frequency discrimination for isolated English vowels

Area of Science:

• Acoustics
• Psychoacoustics
• Speech Perception

Background:

• Understanding auditory perception of speech sounds is crucial for fields like audiology and speech technology.
• Formant frequencies are key acoustic cues for vowel identification.
• Previous research has provided estimates for formant discrimination, but further precision is needed.

Purpose of the Study:

• To precisely measure formant-frequency discrimination thresholds for synthetic English vowels.
• To characterize the resolution of the human auditory system for speech-relevant acoustic parameters.
• To compare these thresholds with existing data for pure tones and complex sounds.

Main Methods:

• Ten synthetic English vowels were created based on a female talker.
• Well-trained subjects performed minimal-stimulus-uncertainty tasks to determine discrimination thresholds.
• Thresholds for increments and decrements in the first (F1) and second (F2) formant frequencies were measured.

Main Results:

• Reliable thresholds were obtained, except when a harmonic interfered with the test formant frequency.
• Formant frequency discrimination thresholds (delta F) followed a piecewise-linear function.
• Thresholds were constant (~14 Hz) in the F1 region (<800 Hz) and increased linearly in the F2 region (~1.5% resolution).

Conclusions:

• The auditory system exhibits frequency-dependent resolution for formant frequencies.
• These findings refine our understanding of auditory processing for vowels.
• The results suggest better-than-previously-estimated auditory resolution in the F2 region for speech sounds.