AUTHOR=Wang Liting , Liu Huan , Zhang Xin , Zhao Shijie , Guo Lei , Han Junwei , Hu Xintao 

TITLE=Exploring Hierarchical Auditory Representation via a Neural Encoding Model

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 16 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.843988

DOI=10.3389/fnins.2022.843988

ISSN=1662-453X

ABSTRACT=By integrating hierarchical feature modeling of auditory information using deep neural networks (DNN), recent functional magnetic resonance imaging (fMRI) encoding studies have revealed the hierarchical auditory representation in the superior temporal gyrus (STG). However, the hierarchical feature modeling of external auditory stimuli is typically derived using supervised deep neural networks (e.g., for classification). The extracted features are then biased towards discriminative features while ignoring general characteristics shared by multiple categories. Consequently, the hierarchy of neural acoustic processing revealed by the encoding model would be biased towards classification. In this study, we proposed an fMRI encoding framework adopting unsupervised hierarchical feature modeling of naturalistic auditory excerpts via DNN, namely, deep convolutional auto-encoder (DCAE) network, to explore the hierarchical auditory representation in the human brain. Our preliminary results showed that the primary auditory cortex was more sensitive to low-level features represented in shallower layers of the DCAE network, while the visual cortex and insula were related to the encoding of high-level features represented in deeper layers. The present study may provide complementary evidence to understand the hierarchical auditory processing in the human brain.