极光城

关于我 About Me

Biography

Research areas, interests and keywords: statistical machine learning, signal processing, pattern recognition, computer vision, data-science, artifical intelligence.

Siyu Chen received bachelor’s degree from the Department of Software Engineering, Harbin Institute of Technology in 2016, and Master’s degree from the Department of Computer Science and Technology, Zhejiang University in 2019.

In 2023, motivated by the prospect that AI will inevitably be encorporated as part of the civillization, he founded Ultrawave Inc. (Ultrawave Technology), the first AI infrastructure exchange platform, allowing every human to manipulate AI like part of one’s physical body, to utilize and profit from AI distributedly as well as collectively.

During 2019-2022, he worked as algorithm researcher in Alibaba Group, focused on enhancement of audio/speech and visual signals.

During 2017-2018, he worked in Netease Inc. leading the research of cutting-edge AI technologies for game application.

Among 2015-2016, he worked in Tencent Inc. specializing in data science technologies and their application in large-scale business decisions.

Researches

NATS: AI aided speech synthesis from neural activity recordings

This research is forcused on re-harnessing speech capability for patients with neurological dysfunctions and celebral lesions. Cooperative insititution: Westlake University.

AliVAD

AliVAD is a extremely light-weight VAD (Voice Activity Detection) algorithm that out-performs Opus and RTC vad algorithm by nearly 20% with much less or equal computational cost. The well designed and pruned AliVAD model is only 100~300KB in size while having 95% accuracy and 0.96% F1-score. A innovative neural network streaming technique is proposed and utilized in AliVAD, allowing very low realtime detection latency down to 1ms.

other research initiated and implemented:

AliDenoise

Speech enhancement for online hosted video streaming. The technique achieves state of the art PSNR, PESQ, STOI and excellent subjective evaluations. This technology adapts to a large variation of devices from servers to enduser phones, and is currently serving most of Alibaba’s multi-media transactions.

YAnimoji (Yet Another Animoji) Facial Performance Transfer

Siyu Chen, Danping Liao Technical manual

YAID: (Yet Another FaceID) Face Identity Verification

Siyu Chen, Jingzhou Chen, Qian Zhong, Yingyu Diao, Minchao Ye

Game Material Super Resolution Based on Generative Adversarial Network

Danping Liao, Siyu Chen

Publications

Multi-stage Stochastic Gradient Method with Momentum Acceleration Signal Processing 2021

Zhijian Luo, Siyu Chen, Yuntao Qian, Yueen Hou Elsevier-Science Direct

Learning Deep Optimizer for Blind Image Deconvolution IJWMIP 2019

Siyu Chen †, Zhijian Luo †, Yuntao Qian world scientific arXiv github

Stochastic Momentum Method with Double Acceleration for Regularized Empirical Risk Minimization IEEE-ACCESS 2019

Zhijian Luo, Siyu Chen, Yuntao Qian

Deep Neural Network Based Hyperspectral Pixel Classification with Factorized Spectral-Spatial Feature Representation IEEE-ACCESS 2019

Jingzhou Chen, Siyu Chen, Peilin Zhou, Yuntao Qian IEEE arXiv github

A Deep Optimization Approach for Image Deconvolution. CoRR abs/1904.07516 (2019)

Zhijian Luo, Siyu Chen, Yuntao Qian

Lane marking detection using convolutional neural network Signal Processing 2019 Best paper

利用卷积神经网络的车道线检测方法 信号处理-2019 优秀论文

Yuntao Qian, Siyu Chen, Zhongchuang Hu

Spectral Image Visualization Using Generative Adversarial Network PRICAI-2018

Siyu Chen, Danping Liao, Yuntao Qian springer dblp arXiv github

Visualization of Hyperspectral Images Using Moving Least Squares ICPR-2018

Danping Liao, Siyu Chen, Yuntao Qian IEEE dblp arXiv

Contacts

Friendly Interactions Welcomed

欢迎 Star 和 Fork 欢迎 添加友链，请在留言板留言即可。

Credits

Some image processing is done using ImageMagick

Besides the theme is provided by TMAIZE, and this website is powered by Jekyll.