Duration weighted Gaussian Mixture Model supervector modeling for robust speaker recognition | IEEE Conference Publication | IEEE Xplore