Speaker recognition through nonstationary vector AR model | IEEE Conference Publication | IEEE Xplore