Computational models of auditory perception from feature extraction to stream segregation and behavior

Area of Science:

• Auditory Neuroscience
• Computational Auditory Processing
• Psychoacoustics

Background:

• Audition involves dynamic processing across multiple timescales, from sub-millisecond brainstem activity to complex musical perception.
• Understanding auditory scene analysis, sound localization, and beat processing is crucial for comprehending auditory perception.
• Existing research often uses synthesized sounds with regular structures, necessitating broader modeling approaches.

Purpose of the Study:

• To review recent advancements in computational models of key auditory processes.
• To integrate findings from behavioral, electrophysiological, and imaging studies.
• To advocate for a neuromechanistic modeling approach for a comprehensive understanding of the auditory system.

Main Methods:

• Review of computational models for sound localization, auditory stream segregation, and beat perception/generation.
• Synthesis of data from diverse experimental methodologies (behavioral, electrophysiological, imaging).
• Focus on integrating knowledge across different levels of description and experimental fields.

Main Results:

• Recent computational models offer insights into dynamic auditory processing.
• Behavioral, electrophysiological, and imaging studies provide rich data on auditory functions.
• Integration of diverse data highlights the complexity of auditory perception.

Conclusions:

• A neuromechanistic modeling approach is essential for bridging disciplines in auditory neuroscience.
• Incorporating plausible neural mechanisms enhances the explanatory power of computational models.
• This integrated approach advances our understanding of the dynamic auditory system.