The correlogram: a visual display of periodicity

Area of Science:

• Speech and Hearing Sciences
• Acoustic Analysis
• Signal Processing

Background:

• Fundamental frequency (F0) extraction is crucial for voice quality analysis.
• Pathological voices with high F0 instability often cause F0 extraction algorithms to fail, compromising data analysis.
• Existing methods struggle with complex voice signals, necessitating new approaches.

Purpose of the Study:

• To present a new method, the correlogram, for visualizing voice signal periodicity.
• To enhance the analysis of pathological voices where conventional F0 extraction fails.
• To enable semi-manual F0 extraction for difficult voice samples.

Main Methods:

• The correlogram is based on waveform-matching techniques used in F0 extraction.
• It displays multiple F0 candidates as dark bands in a 3D plot (time, frequency, correlation).
• The method visualizes periodicity without selecting a single F0 value, showing F0, F0/2, F0/3, etc.

Main Results:

• The correlogram effectively illustrates the complex periodicity in pathological voices.
• It provides valuable insights where standard F0 analysis is inadequate.
• Demonstrated application on synthetic and natural voices, including six pathological cases with various voice qualities.

Conclusions:

• The correlogram is a valuable tool for analyzing pathological voices and understanding voice signal complexity.
• It facilitates semi-manual F0 extraction, expanding the range of analyzable voices.
• This method offers a significant advancement for voice quality assessment in challenging cases.