NIST and NFI-TNO evaluations of automatic speaker recognition