Hybrid DNN-Latent structured SVM acoustic models for continuous speech recognition | IEEE Conference Publication | IEEE Xplore