Development Of Generalizable, Interpretable, And Privacy-Preserving Human-Centric Audio Applications