PhD Position F/M Generalized Multichannel Speech Enhancement with End-to-end Deep Learning [S]

The first stage transforms the input time-domain signal into a 2D representation akin to a spectrogram through a set of learned filters (filterbank);