“…We also recommend additional studies to check whether human infants have filters like the ones with which we endow our biased learner, namely processing only speech, streaming different voices, and processes like our data augmentation. The evidence for a voice activity detection mechanism is very strong: Infants display a stable preference for speech over other types of sounds (Issard, Tsuji and Cristia, 2023), demonstrating, therefore, an early ability to discriminate between speech and non-speech segments. As for the pseudo speaker separation and the pitch augmentation mechanisms, evidence suggests that newborns can discriminate even among unfamiliar voices (Decasper and Prescott, 1984;Floccia et al, 2000), and that this may engage a different brain network than distinguishing between speech sounds (Paquette, Dionne-Dostie, Lassonde and Gallagher, 2018).…”