Original speech and its echo are segregated and separately processed in the human brain

Area of Science:

• Auditory Neuroscience
• Speech Processing
• Psychoacoustics

Background:

• Speech recognition depends on slow temporal modulations (<16 Hz).
• Long-delay echoes in speech reduce temporal modulations but not intelligibility.
• Neural mechanisms underlying echo resilience in speech are unclear.

Purpose of the Study:

• Investigate neural mechanisms for speech recognition despite echo-induced temporal modulation loss.
• Determine if the brain tracks eliminated modulations or segregates speech from echo.
• Examine the role of attention and speech fine structure in echo resilience.

Main Methods:

• Magnetoencephalography (MEG) experiments on human participants.
• Analysis of cortical activity tracking temporal modulations in speech with and without echoes.
• Modeling of neural responses using speech segregation versus whole-echo encoding.

Main Results:

• Cortical activity tracked temporal modulations eliminated by echoes, exceeding basic adaptation.
• Neural responses favored a speech segregation model over a whole-echo model.
• Speech segregation effects persisted with diverted attention but vanished without fine structure cues.

Conclusions:

• The auditory system employs mechanisms like stream segregation to create echo-insensitive speech representations.
• This echo-insensitive representation supports robust speech recognition in reverberant environments.
• Speech fine structure is crucial for auditory stream segregation and echo cancellation.