Hierarchical clustering of mixture models for scaling up speaker recognition