APPARATUS AND METHOD FOR CREATING OPTIMUM ACOUSTIC MODEL

摘要

A second acoustic model unit including a reference foreign language acoustic model, a native speech voice database and a non-native foreign language voice database, a second acoustic model unit including a native language acoustic model of a non-native speaker and a native language voice database of a non- A state similar to each state of the first acoustic model unit is searched by the second acoustic model unit 200 on the basis of the first acoustic model unit, and together with the native speech voice DB and the non-native foreign language voice DB, state voice database of the non-native speaker extracted according to the state of the user and training the reference foreign language acoustic model, a foreign language acoustic model having superior performance can be generated.