Rate-Agnostic Bioacoustics: Heterogeneous Multi-Taxa Classification with Continuous Filterbanks and Fourier Neural Operators
My rating
The paper targets bioacoustic classification across recordings with heterogeneous sampling rates. Its proposed frontend operates directly on each recording at its native sampling rate and is paired with a Fourier Neural Operator backbone that uses progressive temporal-scale fusion, avoiding fixed-rate resampling and associated high-frequency information loss.
Tags: