Minimum trajectory error training for deep neural networks, combined with stacked bottleneck features.

Zhizheng Wu, Simon King

Browse the full Interspeech paper archive.