Jinyu Li

Jinyu Li
PRINCIPAL SCIENCE LEAD
.

I received the Ph.D. degree from Georgia Institute of Technology, Atlanta, in 2008. From 2000 to 2003, I was a Researcher in the Intel China Research Center and Research Manager in iFlytek, China. I joined Microsoft in 2008 and now lead a team to design and improve speech modeling algorithms and technologies that ensure industry state-of-the-art speech recognition accuracy for Microsoft products. My major research interests cover several topics in speech recognition, including deep learning, noise robustness, discriminative training, feature extraction, and machine learning methods.

 

What's New

Our recent robust speech recognition overview paper: Jinyu Li, Li Deng, Yifan Gong, and Reinhold Haeb-Umbach, An Overview of Noise-Robust Automatic Speech Recognition, in IEEE/ACM Transactions on Audio, Speech, and Language Processing, pp. 1-33, 2014.

Selected Honors

  • Microsoft research technical transfer award, 2014
  • Microsoft patent awards, 2008-present
  • Interspeech 2006 Best Student Paper
  • Colonel Oscar P. Cleaver Award (for the highest score on the Ph.D. preliminary exam in ECE, Georgia Institute of Technology, 2004)
  • Guo Moruo Scholarship (the highest honor in USTC)

Thesis

Soft margin estimation for automatic speech recognition,” Georgia Institute of Technology, 2008.

 

Publications

Patent

· Low-footprint Adaptation and Personalization for Deep Neural Network
· Low-rank and Multi-state Deep Neural Network
· exploiting heterogeneous data in deep neural network based speech recognition systems
· Efficient implementation of posterior-based feature with partial distance elimination
· Shared-Hidden-Layer Multilingual Deep Neural Network for Improved Speech Recognition
· Distortion-model-based Noise Robust Algorithm with Scalar Operations
· Unscented Transform with Online Distortion Estimation for HMM Adaptation
· Confidence Calibration in Automatic Speech Recognition Systems
· Model Training from Speech with Imperfect Transcription
· HMM adaptation using a phase-sensitive acoustic distortion model for environment-robust speech recognition
· adaptation of compressed hmm parameters for resource-constrained speech recognition
· high-performance hmm adaptation with joint compensation of additive and convolutive distortions via vector Taylor series