Acoustic voice variation in spontaneous speech.

Area of Science:

• Acoustic phonetics
• Speech processing
• Speaker recognition

Background:

• Previous research indicated similar acoustic voice spaces for male and female talkers in read speech.
• The acoustic voice space describes the range and distribution of voice characteristics.

Purpose of the Study:

• To replicate and extend findings on acoustic voice variation in read speech.
• To compare acoustic voice spaces derived from spontaneous speech versus read speech.
• To investigate the role of fundamental frequency variability in spontaneous speech.

Main Methods:

• Principal component analysis (PCA) applied to acoustic indices of voice quality.
• Analysis of phone conversations from 99/100 talkers previously studied.
• Comparison of acoustic voice spaces from spontaneous and read speech data.

Main Results:

• Acoustic voice spaces from spontaneous speech closely resemble those from read speech.
• Variability in fundamental frequency significantly contributed to acoustic variability in spontaneous speech, unlike in read speech.
• The findings support the robustness of acoustic voice spaces across speech types.

Conclusions:

• Acoustic voice spaces are largely consistent between read and spontaneous speech.
• Fundamental frequency variability is a crucial factor in spontaneous speech analysis.
• These insights can refine prototype models for speaker recognition and discrimination.