A sound direction estimation apparatus includes a sound source model storage unit that stores likelihood of a sound source class in correspondence with a sound feature quantity, a sound signal input unit that receives a sound signal, a sound source identification unit that identifies a sound source class of the sound signal input from the sound signal input unit with reference to the sound source model storage unit based on the sound feature quantity of the sound signal, and a first sound source localization unit that estimates a sound direction of the sound signal of the sound source class identified by the sound source identification unit.