Detrending the Waveforms of Steady-State Vowels

Area of Science:

• Acoustics
• Speech Science
• Signal Processing

Background:

• Steady-state vowels feature quasi-periodic waveforms reflecting vocal tract configuration.
• Pitch-synchronous methods analyze these waveforms using pitch periods as a time scale.
• Existing methods may not fully account for low-frequency content from glottal pulses.

Purpose of the Study:

• To present a simple, pitch-synchronous Bayesian method for estimating vowel formants.
• To generalize existing formant estimation by incorporating glottal pulse effects.
• To efficiently detrend steady-state vowel waveforms.

Main Methods:

• Modeled pitch periods as decaying sinusoids representing formants.
• Added a polynomial trend function to account for low-frequency glottal pulse content.
• Employed a Bayesian approach for parameter estimation.

Main Results:

• Developed an efficient method for detrending steady-state vowel waveforms.
• Successfully separated formant-related sinusoids from the glottal pulse trend.
• Demonstrated a generalized approach to pitch-synchronous formant estimation.

Conclusions:

• The proposed Bayesian method offers improved formant estimation by accounting for glottal pulse artifacts.
• Detrending vowel waveforms is an effective strategy for separating acoustic components.
• This technique provides a more accurate analysis of vocal tract configurations from speech signals.