Pitch-Synchronous Analysis of Human Voice

Area of Science:

• Acoustics and Signal Processing
• Bioengineering
• Speech Science

Background:

• Traditional spectrograms mix pitch and timbre information due to fixed windowing.
• Understanding human voice production requires precise separation of these elements.

Purpose of the Study:

• To develop a pitch-synchronous segmentation method for voice signals.
• To achieve high-resolution visual representations of pitch marks and timbre spectra.
• To gain deeper insights into the human voice production process.

Main Methods:

• Simultaneous acquisition of voice and electroglottograph (EGG) signals.
• Pitch-synchronous segmentation of voice signals based on EGG derivatives or voice signal analysis.
• Fourier analysis applied to individual pitch periods without window functions.
• Development of specialized software for pitch-synchronous spectrograms.

Main Results:

• A novel method cleanly separates pitch and timbre information.
• High-resolution and accurate graphical representations of pitch marks and timbre spectra are generated.
• Modern data substantiates the transient theory of voice production.
• Individual display of timbre spectra and power evolution patterns within each pitch period is possible.

Conclusions:

• Each glottal closing generates a decaying wave, with voice being a superposition of these waves.
• The developed pitch-synchronous method, using EGG data, enables accurate timbre spectrum and power evolution pattern analysis.
• This approach provides a more precise understanding of voice production mechanisms.