A flat direct model for speech recognition | IEEE Conference Publication | IEEE Xplore