Modelling context processing during sentence recognition in noise and reverberation for listeners with and without...

Area of Science:

• Auditory Neuroscience
• Speech Perception
• Human Hearing

Background:

• Context processing is crucial for speech recognition, especially for individuals with hearing loss.
• Understanding how listeners with hearing loss utilize context is vital for developing effective hearing aid strategies.
• Previous models have limitations in accurately quantifying context processing differences.

Purpose of the Study:

• To evaluate three models for quantifying context processing in human speech recognition.
• To test if hearing-impaired listeners rely more on speech context to compensate for reduced auditory input.
• To investigate how context processing differs across various noise types and reverberation conditions.

Main Methods:

• Presented everyday and matrix sentences (varying predictability) to normal-hearing and hearing-impaired listeners.
• Tested sentences in quiet, steady-state noise, and modulated noise, with clean and reverberated speech.
• Utilized bootstrapping for significance testing to compare model performance.

Main Results:

• Only one of the three evaluated context models demonstrated significant differences between normal-hearing and hearing-impaired listeners.
• No significant differences in context processing were observed between stationary and modulated noise conditions.
• Distinguishing between a priori and a posteriori context is necessary to explain speech perception in reverberation.

Conclusions:

• The evaluated models have limitations in capturing context processing nuances in hearing-impaired listeners.
• Speech context compensation strategies may not differ significantly between stationary and modulated noise for these listeners.
• Further research is needed to refine models for understanding context's role in speech recognition under adverse conditions.