Speaker normalization in the perception of Mandarin Chinese tones

Area of Science:

• Phonetics
• Psychoacoustics
• Linguistics

Background:

• Speaker normalization is crucial for speech perception, allowing listeners to adjust for vocal variations.
• Mandarin tones 2 (mid-rising) and 3 (low-falling-rising) are distinguished by acoustic features like turning point and delta F0.
• Listeners utilize fundamental frequency (F0) range to identify speakers and normalize acoustic signals.

Purpose of the Study:

• To investigate how listeners normalize perception of Mandarin tones 2 and 3 based on speaker F0 range.
• To determine the role of acoustic cues (turning point, delta F0) in speaker normalization for Mandarin tones.

Main Methods:

• Production and perception experiments were conducted with two speakers of differing pitch.
• Tone continua were created by varying turning point, delta F0, or both, and appended to natural precursor phrases.
• Listeners identified tones in stimuli presented with high- and low-pitched speaker precursors.

Main Results:

• Identical stimuli were perceived as different tones depending on the speaker precursor (high vs. low pitch).
• Normalization effects were observed, with significant identification shifts occurring.
• Stimuli varying in turning point showed no significant shift, suggesting normalization is dimension-specific.
• Normalization magnitude was greater for stimuli varying only in delta F0 compared to those varying in both dimensions.

Conclusions:

• Listeners normalize Mandarin tones based on speaker F0 range, adjusting their perception of tones 2 and 3.
• Speaker normalization is influenced by the acoustic dimensions of the tone and the precursor phrase.
• Normalization effects are modulated by the degree to which stimuli resemble natural speech variations.