Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model
Keyword(s):