Rhythm-based hierarchical predictive computations support acoustic-semantic transformation in speech processing.

Area of Science:

• Neuroscience
• Computational Auditory Neuroscience
• Speech Processing

Background:

• Human speech comprehension relies on predicting structure and content, potentially involving brain rhythms.
• The neural mechanisms of rhythm-based predictive speech processing remain largely unknown.

Purpose of the Study:

• To propose a neural model, the brain rhythm-based inference model (BRyBI), for speech processing in the auditory cortex.
• To elucidate how endogenous brain rhythms interact within a predictive coding framework to form contextual information for speech.
• To explain human speech recognition performance and experimental findings on brain rhythms during speech perception.

Main Methods:

• Developed the brain rhythm-based inference model (BRyBI) based on predictive coding principles.
• Modeled the interaction of endogenous brain rhythms for spectro-temporal speech parsing and phrasal context formation.
• Compared model predictions with human speech recognition data and experimental observations.

Main Results:

• BRyBI successfully encodes rhythmic processes for parsing speech into phonemes and forming phrasal context.
• The model aligns with human performance patterns in speech recognition tasks.
• BRyBI accounts for variability in brain rhythms observed during speech listening, linked to uncertainty and surprise.

Conclusions:

• Multiscale brain rhythms play a crucial computational role in predictive speech processing.
• The proposed BRyBI offers a plausible neural implementation for rhythm-based speech understanding in the auditory cortex.
• Understanding brain rhythm interactions is key to deciphering the neural basis of speech comprehension.