A scalable approach to using DNN-derived features in GMM-HMM based acoustic modeling for LVCSR.

Zhi-Jie Yan, Qiang Huo, Jian Xu

Browse the full Interspeech paper archive.