This study investigates whether daytime naps facilitate perceptual learning of Cantonese tones and how prior knowledge mediates the consolidation effect. Ninety Mandarin native speakers were pseudo-randomly assigned to either a nap group, who napped for 1.5 h with brain activities recorded, or the non-nap group, who rested for 1.5 h. They were trained with Cantonese contour-level tonal contrasts and level-level tonal contrasts, followed by a tone identification task (trained talker) before the nap manipulation, and were re-tested (trained and novel talkers) after the nap. The results showed that naps facilitated Cantonese tone learning, with the nap group outperforming the non-nap group in the cross-talker perception. The cross-talker perception effect was specific to contour-level tonal contrasts (consistent with prior knowledge) and was associated with increased sleep spindles and slow oscillations. The findings suggest that prior knowledge plays an important role in prioritizing contour-level tonal contrasts for memory consolidation.