Source-normalised-and-weighted LDA for robust speaker recognition using i-vectors | IEEE Conference Publication | IEEE Xplore