Robust speaker recognition based on DNN/i-vectors and speech separation | IEEE Conference Publication | IEEE Xplore