Speech recognition under masking: Age, hearing, and machine learning classification.

Area of Science:

• Audiology
• Machine Learning
• Speech Perception

Background:

• Maskers in auditory research are often categorized as energetic or informational.
• Understanding how age and hearing ability influence speech perception in noise is crucial.

Purpose of the Study:

• To classify maskers into energetic and informational categories using machine learning.
• To investigate the effects of age and hearing impairment on speech reception thresholds (SRTs).
• To compare speech recognition performance across different age groups and hearing abilities under various masker conditions.

Main Methods:

• Machine learning classification techniques were employed to categorize maskers.
• Linear mixed-effects modeling was used to predict signal-to-noise ratios (SNRs).
• Participants included young and older adults with normal hearing, and older adults with hearing impairments (HI).

Main Results:

• Machine learning successfully confirmed the classification of maskers.
• Younger and older normally-hearing adults showed similar performance, especially with informational maskers.
• Older adults with hearing impairments (HI) consistently required higher SNRs across all masker types.

Conclusions:

• Machine learning supports the theoretical categorization of maskers.
• Peripheral hearing status significantly impacts speech recognition more than age.
• Normal-hearing older adults may employ compensatory strategies to match younger adults' performance.