Shengkui Zhao （赵胜奎）
Algorithm Expert - Speech at 阿里巴巴- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
-
English -
-
Mandarin -
Topline Score
Bio
Credentials
-
Machine Learning - Andrew Ng (Stanford University)
Coursera Course CertificatesSep, 2016- Sep, 2024
Experience
-
阿里巴巴
-
1 - 100 Employee
-
Algorithm Expert - Speech
-
Nov 2017 - Present
Leading research and development in speech enhancement (SE), robust speech recognition (ASR), cross-lingual voice conversion (CLVC), and Text-To-Speech (TTS). Built online real-time speech enhancement model for audio-visual conference meeting apps (Dingtalk, Alilang and more). Built first most natural bilingual and code-switching (English-Mandarine) Text-To-Speech (TTS) model Built most natural neural cross-lingual voice conversion model Built speech enhancement model for robust speech recognition in extremely noisy environments Show less
-
-
-
Advanced Digital Sciences Center (ADSC)
-
Singapore
-
Research Services
-
1 - 100 Employee
-
Research Scientist(speech)
-
Nov 2013 - Nov 2017
Roles: Technical team leader; conducting research and development; managing projects; writing proposals and papers; mentoring engineers and interns Project details:1) Project title: Acoustic analysis: acoustic event detection, classification, and analytics.Develop data pipeline that sources, cleanses, segments, transforms and evaluates multi-channel acoustic data. Developed predictive models for detecting human activities such as anomaly sound events.Developed predictive models for classifying sound scenes and environments. Developed algorithms for localizing single or multiple sound sources in indoor or outdoor environments. Develop algorithms for near and distance speech enhancement for robust speech recognition in noisy and reverberant environments using microphone array 2) Project title: Noise-scape: Abating urban noise through a holistic approach of noise monitoring, analytics and controlDesigned a movable microphone array system for acoustic capturing using an electric vehicle.Developed algorithms and visualization tools for displaying locations of sound sources and sound pressure levels in acoustic maps. Engaged various government agencies to discuss project proposals and report project progress. Collaborated with local universities and A*STAR to conduct the NRF Land and Livability National Innovation Challenge project. Show less
-
-
Postdoctoral Fellow(speech)
-
Nov 2010 - Oct 2013
As a postdoctoral fellow at ADSC, I researched on 3D audio enhancement and reproduction, 3D sound source localization, single / multichannel speech enhancement for speech recognition.Project details:1) Project title: Realistic Audio Telepresencing for Entertainment and Meetings (RATEM) Developed algorithms for single / multisource 3D acoustic source localization, and for underdeterrmined sound localization where the number of sensors is smaller than the number of sources; Developed algorithms for 3D audio reproduction for arbitrary arrays and built real-time demo system for capturing and playing back sound in 3D using a headpone; Developed algorithms for noise reduction and dereverberation for microphone array signals and improved speech recognition performance.2) Project title: SSIPO Safe City Test Bed Developed a real-time system to detect fighting incidents in crowded environments Show less
-
-
-
-
Research Fellow
-
Nov 2008 - Nov 2010
Role: conducting research and development on microphone array processing including adaptive beamforming for speech enhancement and the 3-dimensional sound source localization for the social robotics project of A*STAR. During this period, I have been working on 1) Project title: Smart social robotics for machine-human interaction: Olivia robot Developed new microphone array for Olivia robotics 1.2 for 3D speech source localization; Developed algorithms for 3D sound source localization, speech enhancement; and voice activity detection; Collaborated with Institute of Infocomm Research (I2R), A*STAR. 2) Project title: "Who Spoke When" diarization system Developed algorithms for the classification of multiple speakers’ speeches using the direction of arrival (DOA) estimation and collaborated with the Institute for Infocomm Research (I2R), A*STAR. Show less
-
-
Education
-
2005 - 2008
Nanyang Technological University Singapore
Doctor of Philosophy, Computer Engineering (Algorithm) -
2000 - 2004
Nanyang Technological University Singapore
Bachelor of Engineering, Computer Engineering (Programming)
Community
