ByNobleID
    Training log-linear acoustic models in higher-order polynomial feature space for speech recognition | NobleID