Validating a real-time perceptual model predicting distraction caused by audio-on-audio interference.

Area of Science:

• Acoustics
• Psychoacoustics
• Human-Computer Interaction

Background:

• Existing perceptual models for audio distraction are not real-time.
• A new real-time model was developed based on a validated non-real-time model.
• Both models require separate target and interferer signals (non-blind).

Purpose of the Study:

• To validate a newly developed real-time perceptual model for predicting audio-on-audio distraction.
• To compare the accuracy of the real-time model against subjective ratings and an original non-real-time model.
• To assess the suitability of the real-time model for sound-zone evaluation.

Main Methods:

• A validation experiment was conducted comparing model predictions to subjective distraction ratings.
• Listening experiments were performed to gather subjective distraction data.
• Root-mean-squared errors (RMSE) were calculated for speech and music zones for both models.

Main Results:

• The real-time model achieved RMSEs of 10.2% (speech) and 12.6% (music).
• The original model achieved RMSEs of 11.3% (speech) and 11.5% (music).
• The real-time model demonstrated comparable accuracy to the original model.

Conclusions:

• The real-time perceptual model is validated and suitable for sound-zone evaluation.
• The real-time capability is essential for applications like adaptive sound zone assessment.
• The model accurately predicts perceptual distraction from audio-on-audio interference.