A sleep assistance device includes a contactless biometric sensor, a processor, memory, and a speaker. The processor detects a user's sleep state by reading signals from the contactless biometric sensor. The processor may then initiate a wind-down routine upon detecting a sleep-readiness state, including playing relaxing sounds or playing a respiration entrainment sound. The processor may also play noise-masking sounds upon detecting that a user has fallen asleep and seamlessly transition between the sounds played during the wind-down routine and the noise-masking sounds.