Optimised adaptive procedures and analysis methods for conducting speech-in-noise tests

Area of Science:

• Audiology
• Speech perception research
• Psychoacoustics

Background:

• Speech-in-noise testing is essential for audiological evaluations.
• Standardization of test methods is needed for consistent administration.
• Adaptive procedures and analysis methods are key to accurate speech testing.

Purpose of the Study:

• To evaluate the accuracy and reliability of various Bayesian and non-Bayesian adaptive procedures for speech-in-noise testing.
• To compare different data analysis methods for speech intelligibility assessment.
• To identify optimal methods for standardized speech-in-noise test administration.

Main Methods:

• Simulated matrix sentence tests with varying sentence counts (10-50) and target intelligibilities (50%, 75%).
• Modeled listeners with diverse characteristics were used for simulations.
• Assessed seven measurement procedures and three data analysis techniques.

Main Results:

• Estimation of 50% speech intelligibility demonstrated high accuracy and reliability, even with fewer stimuli.
• Achieving 75% intelligibility showed reduced accuracy, necessitating more stimuli.
• Certain Bayesian procedures outperformed others in estimating 75% intelligibility and psychometric function width.

Conclusions:

• A standardized procedure could enhance the consistency of matrix sentence tests across intelligibility levels.
• Candidate adaptive procedures and analysis methods were identified and discussed.
• Findings may extend to other speech materials, with further human participant testing recommended.