😃 About me
I am Dongheon Lee, a ML Engineer at SqueezeBits, formerly a research scientist intern at Meta Reality Labs 
and a postdoctoral researcher at KAIST 
. My research mainly focuses on speech enhancement, separation, and audio generation, including real-time, on-device models. I am interested in AR/VR, and speech recognition. I obtained Ph.D. (2025) and B.S. (2020) at KAIST, advised by Professor Jung-Woo Choi.
For more information, please refer to my CV
📌 Research Interests
- Audio Signal Processing
- Speech Signal Processing
- Microphone Array Processing
- Generative AI (Speech/Audio)
🔥 News
- 2026.06: 🎉 One paper accepted to Interspeech, 2026
- 2025.09: 🎉 One paper accepted to NeurIPS, 2025
- 2025.07: 🎉 1st rank in DCASE 2025 task 4: Spatial Semantic Segmentation of Sound Scenes
- 2025.06: 🎉 Join in Meta Reality Labs as research scientist intern
🏢 Work Experience
- Jan.2026 - now: SqueezeBits
ML Engineer - Jun.2025 - Dec.2025: Meta Reality Labs
Research Scientist Intern, hosted by Juan Azcarreta and Buye Xu, Multichannel speech enhancement (Interspeech 2026) - Jun.2019 - Aug.2019: SK Hynix
Research Intern, DRAN design (1st prize in internship project)
🏫 Education
- Mar.2020 - Feb.2025: Ph.D. Electrical Engineering (KAIST)
Dissertation: “Unified Auditory Scene Analysis and Separation using Dense Frequency-Time Attentive Network”
Advisor: Prof. Jung-Woo Choi - Mar.2016 - Feb.2020: B.S. Electrical Engineering (KAIST)
- Mar.2014 - Feb.2016: Changwon Science Highschool
Early Graduation
📎 Selected Publications
[8] Dongheon Lee, Ashutosh Pandey, Sanjeel Parekh, Daniel Wong, Jacob Donley, Buye Xu, Juan Azcarreta, “Spatial-Magnifier: Spatial upsampling for multichannel speech enhancement,” Annual Conference of International Speech Communication Association (Interspeech) 2026
[7] Dongheon Lee, Younghoo Kwon, Jung-Woo Choi, “DeepASA: An object-oriented one-for-all network for auditory scene analysis,” The 39th Annual Conference on Neural Information Processing Systems (NeurIPS) 2025
[6] Dongheon Lee, Jung-Woo Choi, “DeFT-Mamba: Multichannel universal sound separation and polyphonic audio classification,” International Conference on Audio, Speech, Signal Processing (ICASSP) 2025 (oral)
[5] Dongheon Lee, Jung-Woo Choi, “DeFTAN-II: Efficient multichannel speech enhancement with subgroup processing,” IEEE/ACM Transactions on Audio, Speech, and Language Processing (IEEE/ACM TASLP) 2024
[4] Dongheon Lee, Jung-Woo Choi, “DeFTAN-AA: Array geometry-agnostic multichannel speech enhancement,” Annual Conference of International Speech Communication Association (Interspeech) 2024 (oral)
[3] Dongheon Lee, Dayun Choi, Jung-Woo Choi, “DeFT-AN RT: Real-time multichannel speech enhancement using Dense Frequency Time Attentive Network and non-overlapping synthesis window,” Annual Conference of International Speech Communication Association (Interspeech) 2023
[2] Dongheon Lee, Jung-Woo Choi, “DeFT-AN: Dense Frequency-Time Attentive Network for multichannel speech enhancement,” IEEE Signal Processing Letters (IEEE SPL) 2023
[1] Dongheon Lee, Byeongho Jo, Jung-Woo Choi, “Direction-of-arrival estimation with blind surface impedance compensation for spherical microphone array”, Journal of Acoustical Society of America Express Letters (JASA EL) 2021
📝 Publications (13 publications, 12 first-author)
2026
[C10] Spatial-Magnifier: Spatial upsampling for multichannel speech enhancement
Dongheon Lee, Ashutosh Pandey, Sanjeel Parekh, Daniel Wong, Jacob Donley, Buye Xu, Juan Azcarreta
Annual Conference of International Speech Communication Association (Interspeech) 2026
2025
[C09] DeepASA:An object-oriented one-for-all network for auditory scene analysis
Dongheon Lee, Younghoo Kwon, and Jung-Woo Choi
The 39th Annual Conference on Neural Information Processing Systems (NeurIPS) 2025
Younghoo Kwon*, Dongheon Lee*, Dohwan Kim, and Jung-Woo Choi (*: Equal Contribution)
DCASE Technical Report 1st rank, Winner (DCASE) 2025
[C07] Universal auditory scene analysis model for source separation, event localization, and detection
Dongheon Lee, Jung-Woo Choi
Forum Acusticum Euronoise (EuroNoise) 2025
[C06] DeFT-Mamba: Multichannel universal sound separation and polyphonic audio classification
Dongheon Lee, Jung-Woo Choi
International Conference on Audio, Speech, Signal Processing (ICASSP) 2025 (oral)
2024
[J05, C05] DeFTAN-II: Efficient multichannel speech enhancement with subgroup processing
Dongheon Lee, Jung-Woo Choi
IEEE/ACM Transactions on Audio, Speech, and Language Processing (IEEE/ACM TASLP) 2024
[C04] DeFTAN-AA: Array geometry-agnostic multichannel speech enhancement
Dongheon Lee, Jung-Woo Choi
Annual Conference of International Speech Communication Association (Interspeech) 2024 (oral)
2023
Dongheon Lee, Dayun Choi, and Jung-Woo Choi
Annual Conference of International Speech Communication Association (Interspeech) 2023
[J04, C02] DeFT-AN: Dense Frequency-Time Attentive Network for multichannel speech enhancement
Dongheon Lee, and Jung-Woo Choi
IEEE Signal Processing Letters (IEEE SPL) 2023
2022
[C01] Inter-channel Conv-TasNet for source-agnostic multichannel audio enhancement
Dongheon Lee, Jung-Woo Choi
51st International Congress Exposition on Noise Controal Engineering (Inter-Noise), 2022. (oral)
2021
Zhi-Jun Zhao, Junseong Ahn, Dongheon Lee, et. al.
Nanoscale, 2021
[J02] Inter-channel Conv-TasNet for multichannel speech enhancement
Dongheon Lee, Seongrae Kim, and Jung-Woo Choi
ArXiv, 2021
Dongheon Lee, Byeongho Jo, Jung-Woo Choi
Journal of Acoustical Society of America Express Letters (JASA EL) 2021
Academic Activities
💬 Invited Talks
- 2024 Machine Learning and Big Data, Next-Generation ICT Research Center
- 2023 On-device multichannel speech enhancement system, AICube
- 2023 AI Specialized Training Program, KAIST-Hwaseong Hub
🏆 Awards
-
1st Rank (Winner) of DCASE 2025 Task 4: Spatial Semantic Segmentation of Sound Scenes (2025)
-
Outstanding Teaching Assistant Award, EE488B: Audio Signal Processing (2024)
-
Excellence Paper Award, Acoustical Society of Korea (2021)
-
1st Prize in Internship Project, SK Hynix (2019)
📑 Reviewer
Confernce
-
IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA): 2025 ~
-
International Conference on Audio, Speech, Signal Processing (ICASSP): 2025 ~
-
Proceedings of International Speech Communication Association (Interspeech): 2025 ~
Journal
-
IEEE/ACM Transactions on Audio, Speech, and Language Processing (IEEE/ACM TASLP)
-
IEEE Signal Processing Letters (IEEE SPL)
📏 Teaching Assistant
-
Signals and Systems, Mar. 2021 – Feb. 2024
-
Audio Signal Processing, Feb. 2024 – Jun. 2024
-
Individual Research, Mar. 2021 – Aug. 2024
🎏 Patent
-
Method and device of suppressing outdoor noise by using a microphone array, KR 10-2022-0131773
-
Method and apparatus for array geometry agnostic denoising and dereverberation based on deep learning, KR 10-2024-0063256
-
Voice enhancement method and system in eXtended Reality space for multi-party voice conversation, KR 10-2023-0150037