Data augmentation is a simple and efficient technique to improve the robustness of a speech recognizer when deployed in mismatched training-test conditions. Our paper proposes a new approach for selecting data with respect to similarity of acoustic conditions. The similarity is computed based on a sequence summarizing neural network which extracts vectors containing acoustic summary (e.g. noise and reverberation characteristics) of an utterance. Several configurations of this network and different methods of selecting data using these "summary-vectors" were explored. The results are reported on a mismatched condition using AMI training set with the proposed data selection and CHiME3 test set.
|ジャーナル||Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH|
|出版ステータス||Published - 2016|
|イベント||17th Annual Conference of the International Speech Communication Association, INTERSPEECH 2016 - San Francisco, United States|
継続期間: 2016 9 8 → 2016 9 16
ASJC Scopus subject areas