Language-identification based on cross-language acoustic models and optimised information combination.