Zhengxi Liu

刘正曦

Undergraduate in Sun Yat-sen University

Major in information and computing science (2018-2022)

Interested in speech technology, web development and high performance programming

Full stack developer of MistGPU, and working on mobileboard💪

My Github (270+ followers, 1k+ stars)

contact me

中文版

---

Research Interests

Efficient and high quality neural vocoder

Expressive TTS acoustic model

Educations

09/2018 - 06/2022, Sun Yat-sen University, Guangzhou, China

1. B.S. in Information and Computing Science

Publications

Zhengxi Liu and Yanmin Qian, Basis-MelGAN: Efficient Neural Vocoder Based on Audio Decomposition, to appear in InterSpeech 2021.

1. paper, code
2. Proposed a GAN-based neural vocoder model, which has a novel architecture, using TasNet basis matrix as a part of the model, and have shown this design makes the improvement of inference speed and audio quality.

Awards

1st prize and best performance award on ASC20-21

1. winners site
2. We won the highest performance award and achieved the third best result in the total score. I am mainly responsible for the optimization of AI challenge in our team.

Experiences

Research Intern, AI Platform, HuYa Inc., Guangzhou, 05/2021-07/2021

1. Mentor: Shiyin Kang, Deyi Tuo
2. Focus on efficient neural vocoder.

Research Intern, Lightspeed & Quantum Studios Group, Tencent Inc., Shenzhen, 03/2020-05/2021

1. Mentor: Chengzhu Yu and Shuai Wang
2. Focus on efficient neural vocoder and multi-style TTS model.

Research Intern, Speech team, DMAI Inc., Guangzhou, 07/2019-09/2019

1. Mentor: Xiaodan Liang
2. Focus on multi-speaker TTS model. Explore the application of lifelong learning in multi-speaker TTS.

Research Intern, HCP lab, Sun Yat-sen University, Guangzhou, 10/2018-10/2019

1. Mentor: Xiaodan Liang
2. Acquisition the knowledge of speech synthesis and deep learning.

Skills

Python, C/C++, JavaScript, Pytorch, Docker, MySQL, Flask.