Spectral-normalization filter for subjective analysis of the aging voice

Area of Science:

• Acoustic analysis
• Speech science
• Gerontology

Background:

• Voice quality changes with age, often decreasing quality of life.
• Estimating talker age from voice is key to understanding vocal aging.
• Longitudinal studies offer insights into progressive vocal degeneration, but recording quality is a limitation.

Purpose of the Study:

• To develop a spectral-normalization filter to mitigate recording quality limitations in longitudinal voice analysis.
• To enable more accurate acoustic analysis of voice changes over a 50-year period.
• To investigate age-related vocal function degeneration.

Main Methods:

• Development and application of a novel spectral-normalization filter.
• Analysis of a 50-year longitudinal corpus of an individual's voice recordings (1959-2007).
• Evaluation of filter effectiveness on autospectra and fundamental frequency; preliminary subjective quality assessment.

Main Results:

• The spectral-normalization filter effectively normalized autospectra across recordings from different decades.
• The filter did not significantly affect the fundamental frequency of the voice.
• Preliminary subjective analysis indicated that the filter improved perceptual similarity in recording quality.

Conclusions:

• The developed spectral-normalization filter is effective in addressing technological limitations in historical voice recordings.
• This method facilitates more reliable longitudinal acoustic analysis of voice aging.
• Further research can leverage this technique to better understand age-related vocal changes and their impact on quality of life.