Binary spectral masking selects target components based on local SNR thresholds

Evidence

The spectro-temporal SNR representation was used to decide which components of the target speech signal would be passed on to the final signal. Any local SNR values exceeding the selection criterion resulted in a value of 1 of a binary mask, and thus in selection of the corresponding spectro-temporal components of the target signal in both ears. Any local SNR values that did not pass the criterion resulted in a value of zero and deselection of the corresponding components of the target (Fig. 1 , right column). — Dijk, Pim van, Physiology, Psychoacoustics and…, 1645