Combining cross-stream and time dimensions in phonetic speaker recognition | IEEE Conference Publication | IEEE Xplore