New preprint: Multiscale Temporal Processing Supports Sound Recognition under Causal Constraints.
2026-08-25
In our latest preprint, we introduce the Multiscale Convolutional Recurrent Neural Network (MSCRNN), a brain-inspired architecture designed to recognize sounds online by integrating acoustic information across multiple temporal scales.
We show that multiscale temporal processing substantially improves causal environmental sound recognition while giving rise to progressively richer internal representations that resemble key principles of auditory processing.
Multiscale Temporal Processing Supports Sound Recognition under Causal Constraint (2026). Esposito M, , , ,
