{"606842":{"#nid":"606842","#data":{"type":"event","title":"Ph.D. Dissertation Defense - Zhong Meng","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle\u003C\/strong\u003E\u003Cem\u003E:\u0026nbsp; \u003C\/em\u003E\u003Cem\u003EDiscriminative and Adaptive Training for Robust Speech Recognition and Understanding\u003C\/em\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003ECommittee:\u003C\/strong\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Biing-Hwang Juang, ECE, Chair , Advisor\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Chin-Hui Lee, ECE\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Elliott Moore, ECE\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. James McClellan, ECE\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Yao Xie, ISyE\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003EAbstract:\u003C\/strong\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003ERobust automatic speech recognition (ASR) and understanding (ASU) under noisy conditions remains to be a challenging problem even with the advances of deep learning.\u0026nbsp;To achieve robust ASU, two discriminative training objectives are proposed for keyword spotting and topic classification: (1) To accurately recognize the semantically important keywords, the non-uniform error cost minimum classification error training of DNN and BLSTM acoustic models is proposed to minimize the recognition errors of only the keywords. (2)\u0026nbsp;To compensate for the mismatched objectives of speech recognition and understanding, minimum semantic error cost training of the BLSTM acoustic model is proposed to generate semantically accurate lattices for topic classification.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EFurther, to expand the application of the ASU system to various conditions,\u0026nbsp;four adaptive training approaches are proposed to\u0026nbsp;improve the robustness of the ASR under different conditions: (1)\u0026nbsp;To suppress the effect of inter-speaker variability on speaker-independent DNN acoustic\u0026nbsp;model, speaker-invariant training is proposed to learn a deep representation in the DNN that is both senone-discriminative and speaker-invariant through adversarial multi-task training\u0026nbsp;(2)\u0026nbsp;To achieve condition-robust unsupervised adaptation with parallel data, adversarial teacher-student learning is proposed to suppress multiple factors of condition variability\u0026nbsp;in the procedure of knowledge transfer from a well-trained source domain LSTM acoustic model to the target domain.\u0026nbsp;(3)\u0026nbsp;To further improve the adversarial learning for unsupervised adaptation with unparallel data, domain separation networks are used to enhance the domain-invariance of the\u0026nbsp;senone-discriminative deep representation by explicitly modeling the private component that\u0026nbsp;is unique to each domain. (4)\u0026nbsp;To achieve robust far-field ASR, an LSTM adaptive beamforming network is proposed to estimate the real-time beamforming filter coefficients to cope with non-stationary environmental noise and dynamic nature of source and microphones positions.\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"Discriminative and Adaptive Training for Robust Speech Recognition and Understanding "}],"uid":"28475","created_gmt":"2018-06-07 20:18:52","changed_gmt":"2018-06-07 20:18:52","author":"Daniela Staiculescu","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2018-06-22T11:00:00-04:00","event_time_end":"2018-06-22T13:00:00-04:00","event_time_end_last":"2018-06-22T13:00:00-04:00","gmt_time_start":"2018-06-22 15:00:00","gmt_time_end":"2018-06-22 17:00:00","gmt_time_end_last":"2018-06-22 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"434381","name":"ECE Ph.D. Dissertation Defenses"}],"categories":[],"keywords":[{"id":"100811","name":"Phd Defense"},{"id":"1808","name":"graduate students"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}