Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Neural Representations of Speech and Music under Temporal Compression

Zhengwu Ma¹, Qimiao Gao¹, Jixing Li¹; ¹City University of Hong Kong

Presenter: Zhengwu Ma

Both speech and music rely on structured temporal sequences, yet how neural representations of these sequences adapt or degrade under temporal compression remains less understood. Here, we address this question using an interpretable convolutional neural network (CNN) model previously validated for modeling human auditory processing. We recorded high-density electroencephalography (EEG) while participants listened to a 2-minute AI-synthesized Mandarin song presented at four playback speeds. For each condition, we extracted syllable-level features from the CNN and aligned them with EEG time courses to assess neural tracking of linguistic and musical information. We find that at slower playback rates, both speech and music features exhibited comparable activity in left temporal regions. In contrast, at higher compression rates, only musical features remained significantly correlated with EEG responses, primarily in the right temporal regions. These findings suggest that linguistic and musical processing rely on partially dissociable neural mechanisms that differ in their tolerance to temporal compression, with musical representations showing greater robustness at faster rates.

Topic Area: Auditory, Speech & Language Processing