Yuanning Li

The paper investigates how human temporal cortex integrates facial and acoustic cues during speech, revealing distinct frequency-dependent integration regimes in different regions.

The paper analyzes how linguistic context influences visual cortex representations using fMRI data from a caption-image matching task, identifying two distinct voxel populations with different response patterns.

This study shows that attention improves population codes in the human visual cortex primarily through tuning changes, not correlation changes.

This paper introduces a dual-path framework for reconstructing high-fidelity speech from neural recordings, achieving high intelligibility with limited ECoG data.

This paper investigates the neural coding of articulatory dynamics during speech imagery and production, revealing a hierarchical spatiotemporal organization and developing a model to decode speech from neural signals.

This study compares how speech and language DNN models align with neural activity during speech processing, finding shared components and distinct spatiotemporal dynamics.

This paper proposes a pre-trained encoder-decoder framework to re-synthesize natural speech from neural recordings, achieving a phoneme error rate of 28.6% and 71.6% word recognition.

This paper develops a brain-to-text framework for decoding natural Mandarin tonal sentences from neural recordings, achieving higher accuracy than previous methods.

This study uses deep neural networks to model neural coding in the human auditory pathway and finds that DNN representations align with biological auditory processing.

The paper shows that neural activity in auditory cortex encodes temporal context in speech through low-dimensional latent states, with timing cues capturing more variance than phonetic features.

This paper improves voxel-wise encoding models by integrating DNN features and brain network information, achieving better prediction of whole-brain neural activity during naturalistic video perception.

The paper describes an ensemble of deep neural networks that predict neural responses to naturalistic videos using multiple input perspectives, achieving state-of-the-art performance.

The paper shows endogenous neural activity modulates neural tuning and perception in a circuit- and stimulus-specific manner using intracranial recordings from epilepsy patients.

SCOTS maps brain function from neural time series by integrating spatio-temporal dependence through dimensionality reduction and classification, revealing category-specific neural activation patterns.

This study shows that the left mid-ventral temporal cortex (lmVTC) individuates real words through interactions with anterior temporal and visual regions, distinguishing them from pseudowords and other stimuli.

The paper investigates the role of posterior and mid-fusiform regions in facial expression processing using intracranial recordings and multivariate analysis, revealing distinct temporal dynamics in their contribution to expression perception.

This paper introduces MCPA to decode information in neural interactions, showing it captures image individuation and face recognition through distributed brain circuits.