ISCA Archive Interspeech 2023
ISCA Archive Interspeech 2023

The effect of masking noise on listeners’ spectral tilt preferences

Olympia Simantiraki, Yannis Pantazis, Martin Cooke

Speech enhancement algorithms often focus on optimising intelligibility while neglecting other aspects of speech such as naturalness, quality and listening effort which may affect a listener's experience. This paper investigates the impact of spectral tilt on listeners' preferences, using a new corpus of Greek utterances. Participants adjusted spectral tilt with real-time feedback to select their preferred tilt in quiet and in the presence of speech-shaped noise at eight signal-to-noise ratios. Listeners displayed distinct preferences, with a tendency to select flatter tilts with increasing noise. Preferences were not random even for constant intelligibility, indicating that their adjustments were influenced by factors beyond the need to maintain comprehensibility. These findings have the potential to inform the design of speech enhancement algorithms that jointly optimise intelligibility and a listener's overall experience.