The relations between area functions and the acoustic signal

Area of Science:

• Acoustic theory of speech production
• Speech signal processing
• Vocal tract modeling

Background:

• Deriving speech waves from area functions and predicting area functions from speech waves are fundamental challenges.
• Deviations from ideal resonator theory, including boundary conditions and source-filter interactions, complicate these processes.

Purpose of the Study:

• To investigate methods for deriving speech wave data from area function specifications and the inverse.
• To address challenges in vocal tract modeling and speech signal analysis.

Main Methods:

• Discussion of deviations from ideal resonator theory.
• Application of perturbation theory for vocal scaling.
• Analysis of shortcomings in inverse transform methods.
• Emphasis on lossy transmission line theory over linear prediction.
• Illustration of bandwidth usage for ambiguity reduction in simple models.

Main Results:

• Lossy transmission line theory shows advantages over standard linear prediction.
• Bandwidth data, when combined with formant frequencies and vocal tract constraints, proves effective.
• Perturbation theory is relevant for male-female vocal scaling.

Conclusions:

• A combination of limited bandwidth data, formant frequencies, and model-related vocal tract constraints is optimal for speech analysis.
• Lossy transmission line theory offers a superior approach to inverse problems in speech production.