Direct models for phoneme recognition | IEEE Conference Publication | IEEE Xplore