Effects of spectral modulation filtering on vowel identification

Area of Science:

• Auditory Perception
• Speech Science
• Acoustic Phonetics

Background:

• Vowel identification relies on complex spectral information.
• Spectral modulation is a key feature of speech acoustics.
• Understanding spectral cues is crucial for speech perception research.

Purpose of the Study:

• To quantify the impact of high-pass spectral modulation filtering on American-English vowel identification.
• To determine the critical spectral modulation frequencies for vowel perception.
• To investigate how spectral manipulations affect acoustic properties of vowels.

Main Methods:

• Applied high-pass filtering at various cutoff frequencies (0.0-2.0 cycles/octave) to naturally spoken American-English vowels.
• Measured vowel identification accuracy for five normal-hearing listeners at 70 dB SPL.
• Conducted acoustic analyses of spectral shape, contrast, and tilt.

Main Results:

• Vowel identification performance decreased as spectral modulation cutoff frequency increased.
• Back vowels showed greater identification degradation than front or central vowels.
• Spectral modulation filtering led to a more crowded vowel space, reduced spectral contrast, and decreased spectral tilt.

Conclusions:

• Spectral cues crucial for vowel identification are primarily below 2 cycles/octave.
• Global spectral feature changes significantly reduce vowel identification accuracy.
• Findings highlight the interplay between spectral shape, spectral smearing, and speech perception.