A simulation framework for auditory discrimination experiments: Revealing the importance of across-frequency...

Area of Science:

• Auditory Perception
• Signal Processing
• Acoustics

Background:

• Auditory discrimination experiments are crucial for understanding hearing capabilities.
• Existing models often focus on speech recognition thresholds, limiting their application to psychoacoustic tasks.
• A validated simulation framework is needed to predict various auditory thresholds.

Purpose of the Study:

• To extend a simulation framework for predicting psychoacoustic thresholds alongside speech recognition thresholds.
• To evaluate the efficacy of different auditory-inspired feature sets in simulating auditory discrimination experiments.
• To assess the performance of the framework across diverse psychoacoustic and speech recognition tasks.

Main Methods:

• A simulation framework, adapted from previous speech recognition threshold prediction models, was utilized.
• The framework was applied to auditory discrimination experiments including tone-in-noise masking, spectral masking, and the German Matrix sentence test.
• Feature sets evaluated included Gabor filter banks, Mel-frequency cepstral coefficients, Mel-spectrograms, and the Dau Perception Model.

Main Results:

• The extended framework successfully simulated all tested auditory experiments using a common parameter set.
• Simulated reference-free thresholds showed agreement with empirical data, validating the predictive capability of the framework.
• Across-frequency processing was identified as critical for accurately modeling speech reception thresholds in modulated noise.

Conclusions:

• The proposed framework provides a versatile tool for simulating a range of auditory discrimination experiments.
• The study demonstrates the suitability of specific auditory-inspired feature sets for predicting psychoacoustic and speech thresholds.
• The findings underscore the significance of across-frequency processing in auditory perception models.