The analysis on speaker timbre perception and acoustic parameters under whispered speech

This study focuses on Mandarin Chinese whispered speech, designing an experiment to distinguish speaker timbre perception using vowels (a, i, u). It then extracts the formants (F1F3) and compares them with normal speech, exploring the characteristics of speaker timbre perception and acoustic parameters under whispered speech conditions. The results show that: (1) From the perspective of human whispered perception, the accuracy of distinguishing speaker timbre perception in whispered speech is about 70%, with listeners more likely to confuse different speakers for the same one. (2) Acoustically, compared to normal speech, whispered speech has significantly higher F2 and F1 values, though F3 values show little difference. In whispered speech, the F2/F1 resonance peak ratio for the vowel i and the F3/F2 resonance peak ratio for the vowel u both drop significantly; the area of the vowel triangle under whispered speech shrinks and shifts leftward and downward. (3) Statistically, under whispered speech conditions, the resonance peaks show a greater relative distance, indicating a higher confusion rate in distinguishing speaker timbre perception.