. . . "should make it usable for applications such as speaker recognition," .