Learning statistically efficient features for speaker recognition | IEEE Conference Publication | IEEE Xplore