To provide a learning device capable of setting a noise suppression parameter set based on a sound collection condition.SOLUTION: A voice database that stores a correct answer corresponding to an audio file and a sound collection condition as a first entry in association with each other is used, and a feature quantity, a sound collection condition, a noise suppression parameter set, and accuracy information are associated with each other A speech recognition result database creating unit that searches the speech recognition result database as a second entry and a speech recognition result database based on the designated sound collection condition and stores the retrieved second entry as a group based on the feature amount A noise suppression parameter set switching rule learning unit for grouping the noise suppression parameter set according to the division standard and storing the optimum noise suppression parameter set, the grouping reference and the sound collection condition in the switching rule database as the third entry in association with each other.BACKGROUND OF THE INVENTION
展开▼