Acoustic parameters in human speaker recognition

Area of Science:

• Acoustic Phonetics
• Speech Perception
• Speaker Recognition

Background:

• Speaker identification is crucial for forensic and security applications.
• Familiar voice recognition relies on various acoustic and prosodic cues.
• Individual voice characteristics significantly influence the perception of speech.

Purpose of the Study:

• To investigate the perceptual importance of F0 height, F0 contour, and speech rhythm in identifying familiar female speakers.
• To determine the relative contribution of these factors to speaker recognition accuracy.

Main Methods:

• Conducted four speaker identification tests using acoustic recordings of reiterant "ma" syllables.
• Employed five female speakers known to the listeners.
• Systematically varied F0 height, F0 contour, and speech rhythm to assess their perceptual relevance.

Main Results:

• F0 height was a highly relevant cue for speakers with typically low or high voices.
• F0 contour played a secondary role in speaker identification for all speakers.
• Speech rhythm had a small but consistent influence on recognition rates.
• Global spectral information alone accounted for approximately 50% of recognition scores.

Conclusions:

• The relevance of perceptual cues in familiar voice recognition is not fixed but depends on speaker-specific voice characteristics.
• F0 height is a primary cue for identifying speakers with distinct pitch ranges.
• Prosodic features like F0 contour and speech rhythm contribute to, but are secondary to, pitch in familiar speaker identification.