Auditory "bubbles": Efficient classification of the spectrotemporal modulations essential for speech intelligibility

Area of Science:

• Auditory perception
• Speech processing
• Psychoacoustics

Background:

• Speech intelligibility relies on intact spectrotemporal patterns.
• The speech modulation power spectrum (MPS) represents energy across temporal and spectral modulation rates.
• Understanding which MPS regions are critical for intelligibility is key.

Purpose of the Study:

• To develop and validate a psychophysical method for identifying essential MPS regions for speech intelligibility.
• To create "perceptual maps" (classification images) of speech intelligibility.
• To investigate how time compression affects these perceptual maps.

Main Methods:

• Utilized a "bubbles" technique adapted from visual perception research.
• Randomly filtered the speech signal by removing portions of the MPS.
• Related filter patterns to listener performance (keyword identification) to generate classification images (CImgs).

Main Results:

• Classification images (CImgs) were reliably estimated for individual listeners.
• CImgs reflected changes in spectrotemporal modulation energy due to time compression.
• Perceptual maps showed deviations from the physical stimulus energy, indicating non-linear processing.

Conclusions:

• The "bubbles" method effectively generates perceptual maps of speech intelligibility.
• These maps highlight the spectrotemporal regions crucial for understanding auditory sentences.
• Auditory perception of speech involves complex processing beyond simple energy detection.