Everyone should learn to tell the difference between good and bad, and between easy and difficult.
每个人都应该学会分辨好与坏、易与难。
It's absurd how often people confuse capability, pressure, hard work, and luck as if they were the same thing.
非常可笑，能力、压力、努力、幸运，有的时候大家竟然会误以为是同一种东西。
I dislike—and sometimes look down upon—those who are full of ideas yet fail to appreciate the difficulty of real implementation.
我不喜欢，有时甚至瞧不起，那些总是有很多想法，却不懂得欣赏工程实现背后复杂性的人。
By “design and implement,” I mean thinking, planning, architecting, coding, testing, deploying, and debugging.
当我说“设计并实现”，我指的是思考、规划、架构、编码、测试、部署和调试。
About
- I am Junqi Deng. I am a researcher, an engineer, and a musician.
- I got my PhD degree in computer engineering at The University of Hong Kong.
- I was deeply involved in the Music Information Retrieval community.
- I design and implement computer systems for music applications.
- I play jazz guitar.
Work
- Quwan Technology (2024.2 - present)
I design and implement Tempolor's first LLM based lyrics conditioned music generation model, which may be the first in China to achieve a similar effect as Suno's, allowing arbitrary lyrics and various other music related descriptions as generative conditions. - SoMuse AI (2022.9 - 2023.8, parttime)
I lead the design and implementation of a singing voice synthesis system（歌声合成系统）for the SoMuse AI App. - Alibaba Taobao Multimedia Technology Unit (2020.7 - 2022.6)
I lead the design and implementation of video music retrieval systems in Taobao GuangGuang（淘宝逛逛里面的智能配乐系统）.
I design and implement mobile audio units and digital audio workstations for Taobao Live and Qinpai（淘宝直播和亲拍里面的音频处理模块）. - Alibaba XLab (2019.7 - 2020.7)
I design and implement the core functionalities in the first version (can be simply understood as a mobile version of Band-in-a-box) of JingBand（鲸乐队）in the JingMing（鲸鸣） App.
I lead the design and implementation of HeyHo Autorap（嘿猴说唱） in the HeyHo App （嘿猴短视频）with a group of three people.
I contribute significantly to the design, implementation and maintenance of "Rhyme King" （押韵帝）, a wechat mini program that helps rappers to find better words or phrases to rhyme. - Alibaba Digital Media and Entertainment Group (at Xiami Music) (2017.12 - 2019.7)
I design and implement Xiami's audio fingerprinting system.（虾米音频指纹系统）
I design and implement several auto-accompaniment systems（自动伴奏系统） for the Music Exploration Lab（探乐实验室） in Xiami App.
- Tencent Music Entertainment Group (at Kugou Music) (2016.10 - 2017.12)
I designed and implemented an auto-accompaniment system,
a music genre classification system for both Chinese and non-Chinese music,
and a word2vec based music recommendation scheme.
All of these had been put into real production.
Education
- Ph.D., Computer Engineering, The University of Hong Kong (2012 - 2016)
- M.E., Integrated Circuit Engineering, Sun Yat-sen University (2010 - 2012)
- B.E., Microelectronics, Sun Yat-sen University (2006 - 2010)
Book Chapters
- 音频音乐与计算机的交融——音频音乐技术，复旦大学出版社，2019年12月（第四章，第七章）
Publications
- Deng, J., Kwok, Y. K., Large Vocabulary Automatic Chord Estimation Using Bidirectional Long Short-Term Memory Recurrent Neural Network with Even Chance Training, Journal of New Music Research (JNMR), Volume 47, Issue 1, 2018
- Deng, J., Kwok, Y. K., Large Vocabulary Automatic Chord Estimation Using Deep Neural Nets: Design Framework, System Variations and Limitations, arXiv preprint arXiv:1709.07153.
- Deng, J., Kwok, Y. K., Large Vocabulary Automatic Chord Estimation with an Even Chance Training Scheme, In Proceedings of the 18th International Society for Music Information Retrieval Conference, Suzhou, China, 2017 (ISMIR 2017)
- Deng, J., Large Vocabulary Automatic Chord Estimation from Audio Using Deep Learning Approaches. PhD thesis, Department of Electrical and Electronic Engineering, The University of Hong Kong, 2016
- Deng, J., Kwok, Y. K., A Hybrid Gaussian-HMM and Deep Learning Approach For SeventhsBass Automatic Chord Estimation, In Proceedings of the 17th International Society for Music Information Retrieval Conference, New York City, USA, 2016 (ISMIR 2016) (Oral)
- Deng, J., Kwok, Y. K., Automatic Chord Estimation on SeventhsBass Chord Vocabulary Using Deep Neural Network, In Proceedings of the 41st International Conference on Acoustics, Speech, and Signal Processing, Shanghai, China, 2016 (ICASSP 2016)
- Hu, X., Deng, J., ..., Kwok, Y. K., SAfeDJ: A crowd-cloud codesign approach to situation-aware music delivery for drivers, ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 2015
- Hu, X., Deng, J., ..., Fels, S., SAfeDJ community: situation-aware in-car music delivery for safe driving, In Proceedings of the 20th annual international conference on Mobile computing and networking, Maui, Hawaii, 2014 (MobiCom 2014)
- Deng, J., Lau, F. C. M., Kwok, Y. K., ArmKeyBoard: A Mobile Keyboard Instrument Based on Chord-Scale System and Tonal Hierarchy, In Proceedings of 40th International Computer Music Conference, Athens, Greece, 2014 (ICMC 2014) (Oral)
- Deng, J., Lau, F. C. M., Ng, H. C., Kwok, Y. K., Chen, H. K., Liu, Y. H., WIJAM: A Mobile Collaborative Improvisation Platform under Master-Players Paradigm, In Proceedings of the 2014 International Conference on New Interfaces for Musical Expression, London, UK, 2014 (NIME 2014)
Projects
- Deep Learning Approaches For Large Vocabulary Automatic Chord Estimation (PhD)
- SAfeDJ (Android) - A music recommendation app for drivers. Take into account the context. Driving alone? Driving with friends? Mood. Maybe others. App contest finalist, the 20th Annual International Conference on Mobile Computing and Networking (MOBICOM 2014)
- WIJAM (iOS) - An application for people to constitute an ad-hoc group and jam along with a music master, whose job is to provide a backing music track and direct the musical flow, App contest finalist, the 19th Annual International Conference on Mobile Computing and Networking (MOBICOM 2013)
- Adding a Prefetching Pass to LLVM Polly – Generating Load-balanced and Coarse-grain Loop Pipeline Code for More Parallelism and Data Locality, Google Summer of Code 2012
- Shang - A High-Level-Synthesis Compiler based on LLVM (optimization modules for auto parallelization and data locality), Master Thesis, 2012
- FPGA based Audio Effect Unit (Reverb, Delay and Distortion) - Course Project 2011
- DSP based Audio Effect Unit (Reverb, Delay and Distortion) - Bachelor Thesis, 2010
Reviewer
- International Society for Music Information Retrieval Conference (2016, 2017, 2018, 2019, 2020, 2021)
- Journal of Audio Engineering Society (Special Issue on Intelligence Audio Processing, Semantics and Interaction, 2016)
- IEEE Transactions on Audio, Speech and Language Processing (2016)
Teaching Assistant
- CCST9015 (Electronic Technologies in Everyday Life, HKU, 2014, 2015)
- CCST9003 (Everyday Computing and the Internet, HKU, 2013, 2014, 2015)
- ELEC2401 (Computer Architecture, HKU, 2013)
- Electronic Device and Circuit Theory (SYSU, 2010)
- Digital Signal Processing Advanced Experiment (SYSU, 2010)
Skills/Hobbies
- Programming:
Now I use Python, C++, C, Swift, SQL most often;
I also do creative programming and audio programming with SuperCollier, Max/MSP occasionally;
I did a bunch of others things in the very early days, including web programming, iOS/Android development, and hardware programming: Objective-C, Java, JavaScript, Html, PHP, Verilog, Labview - Music:
I mainly play guitar and play jazz specifically.
I am a singer-songwriter, and do music productions, arrangements from time to time.
Blog
RSS
Apps
- JustMetronome on iOS - A Simple Professional Metronome
- JustTimer on iOS - A minimalist pomodoro timer, just tap and swipe around the screen, you know how to use it instantly.
- JustTimer on Mac - Just a Pomodoro timer with a simple menu bar hourglass, an optional window interface and easy to use keyboard shortcuts.
