Elia Formisano

The paper introduces a database of 1,377 natural sounds with behavioural norms for use in auditory cognition and neuroscience research.

This paper introduces a brain-inspired neural network model to test if multiscale temporal processing improves sound recognition under causal constraints, finding it does with limited benefit when future information is available.

The paper identifies novel sound features in the human auditory cortex using deep sound synthesis and neuroimaging, showing that unnatural sounds can elicit natural categorical responses.

This study uses fMRI to show that primary and non-primary auditory cortex process speech in noisy environments differently, with one tracking both relevant and non-relevant speech and the other selectively representing relevant speech.

This paper improves sound recognition by incorporating semantic information into deep neural networks, showing better alignment with human perception than traditional categorical models.

This paper compares CNN speech recognition layers to human auditory fMRI activity and finds no shared acoustic feature hierarchy, with all regions most similar to the first fully-connected layer.

The paper investigates how cortical responses to spoken words are encoded with high temporal fidelity, showing that time-locking to speech's spectrotemporal features is crucial for linguistic representation.

This paper introduces a new analytical method for estimating the noise ceiling in fMRI data, comparing it to existing methods and showing its advantages for real data.