Adaptation to Noise in Human Speech Recognition Depends on Noise-Level Statistics and Fast Dynamic-Range Compression.

Area of Science:

• Auditory Neuroscience
• Speech Perception
• Psychoacoustics

Background:

• Human hearing demonstrates adaptation to background noise, enhancing speech recognition.
• This adaptation is hypothesized to involve neural dynamic range shifts improving neural encoding of speech cues.
• The role of stimulus statistics in neural adaptation for speech recognition remains to be fully elucidated.

Purpose of the Study:

• To investigate the importance of statistical adaptation for improving speech recognition in noisy environments.
• To compare the effects of constant versus fluctuating noise precursors on speech recognition adaptation.
• To determine the role of auditory peripheral compression and the medial olivocochlear reflex in noise adaptation.

Main Methods:

• Compared recognition of noise-masked words with and without adapting noise precursors.
• Precursors varied in level: constant or fluctuating every 50 ms based on statistical distributions.
• Measured adaptation as recognition improvement and assessed medial olivocochlear reflex activation and amplitude compression effects.

Main Results:

• Adaptation was largest for constant-level noise precursors and absent for highly fluctuating precursors, despite equal mean levels.
• Both stationary and fluctuating noises activated the medial olivocochlear reflex.
• Fast amplitude compression of fluctuating precursors yielded adaptation comparable to constant-level precursors.

Conclusions:

• Noise adaptation in speech recognition is likely mediated by neural dynamic range adaptation to the most frequent noise level.
• Auditory peripheral compression, by reducing noise level fluctuations, appears to facilitate noise adaptation.
• The medial olivocochlear reflex does not appear to be the primary mediator of this noise adaptation phenomenon.