Experience

• 

  Tencent Cloud

  • Beijing, China

  • Senior Member Research Staff

    • Oct 2021 - Present
    • Beijing, China

    Research lead for media intelligence in Tencent Meeting

• 

  Tencent

  • Research Lead

    • Nov 2017 - Present

    Research Lead for Tencent TEG info security Beijing Group

  • Senior Staff Researcher

    • Oct 2016 - Present

    1. Robust Large Vocabulary Speech recognition for "lack of resource" language and web channels for all Tencent product, including Wechat, QQ(minority languages and strong accent);2. Speaker and language verification & identification;3. Fuzzy keyword spotting&wakeup algorithms and Query by Example related STD (Spoken terms detection) tasks for "lack of resource" language4. Audo Finger print search

  • Research Expert

    • Nov 2018 - Oct 2021

    Focus ASR, KWS, OCR tech for low resource languages

• 

  Dolby Laboratories

  • United States
  • Computers and Electronics Manufacturing
  • 700 & Above Employee

  • Staff Researcher

    • Nov 2014 - Oct 2016

    1. Robust, far voice C-D-RNN based speech recognition by both microphone array optimization and sophisticated audio scene analysis (Lead member)2. Fuzzy keyword spotting & rendering & playback using spatial ASR (Lead member)3. Spatial sound-field resilient transmission algorithm design (Lead member)4. In service morning VOIP performance (Lead member)5. P.800, P.835, P.1131 Subjective listening test design and implementation

  • Senior Reserach Engineer

    • Jul 2011 - Nov 2014

    1. Frontend speech processing algorithms for spatial and mono VoIP and recognition in Dolby Voice, incl VAD, NS, AEC, De-reverb, AGC, DRC, Limiter, etc2. Advanced objective performance metric design for above speech frontend algorithms in Dolby Voice; (Lead member)3. Voice codec, particularly algorithm design for packet loss concealment and forward error correction algorithms in codecs’ network resiliency; (Lead member)4. In service monitoring VoIP network performance for alarms and diagnosis using machine intelligence; (Key member)5. Subjective listening test and results analysis (MUSHRA, ITU.P.XXX, BS1116, non-parametrical test etc), benchmarking existing state-of-art frontend system in commercial product;

• 

  Chinese Academy of Sciences

  • Assistant Researcher

    • Sep 2009 - Jun 2011

    1. Mutimedia information retrieval especially query by humming, music classifi-cation and audio fingerprint search;2. Speech signal analysis & process & recognition;3. Computer aided language learning which designs intelligent systems that can better simulate teachers in understanding the proficiency and intelligibility performance of L2 learner's speech, provides efficient and effective tools by which children can learn more, feel more. Estimation, Diagnosis of Goodness Of Pronunciation, Goodness Of Fluency, Goodness of pRosody. That's the two level, three dimensional structures in Computer Aided Language Learning that I'am always thinking about.