AudioMOS Challenge · ASRU 2025
- 1st Place Track 3: MOS prediction for speech in high sampling frequencies Paper: HighRateMOS
News
-
2026
- We successfully hosted the first Real-World Audio-Visual Speech Enhancement (AVSE) Challenge. We look forward to seeing you at ISCSLP 2026 in Penang, Malaysia! Website · Leaderboard · Baseline
- Two papers accepted to IEEE SLT 2026: correlation-guided encoder selection for multi-encoder audio LLMs, and a holistic evaluation of auditory knowledge in LLM backbones.
- RRP-Voice (first author), a longitudinal voice dataset and benchmark for recurrent respiratory papillomatosis detection, was accepted to APSIPA ASC 2026.
- TaigiSpeech (core contributor), a real-world Taiwanese speech-intent dataset, was accepted to Interspeech 2026 as a long paper.
- MOS-Bias (first author), which uncovers hidden gender bias in speech quality assessment, was accepted to Interspeech 2026.
- Game-Time accepted to ICASSP 2026.
- DeSTA2.5-Audio and CodecFake+ published in IEEE TASLP.
-
2025
- HighRateMOS (first author), sampling-rate-aware speech quality assessment, accepted to IEEE ASRU 2025.
- Two papers at the Interspeech 2025 AVSEC workshop, including BAV-MossFormer2 (first author) for binaural audio-visual enhancement.
- Joint spectral–spatial Mamba for multichannel speech enhancement (first author) at ICASSP 2025.
- Dynamic-SUPERB Phase-2 at ICLR 2025.
-
2024
- EMO-Codec (first author) at APSIPA ASC 2024.
- DFADD at IEEE SLT 2024.
- Robust audio-visual speech enhancement (first author) at O-COCOSDA 2024.
Publications
Honors
Competitions & awards
-
2025
AVSEC-4 Challenge · Interspeech 2025
- 1st Place Single-Channel Track
- 2nd Place Dual-Channel Track
Service
Reviewer
- Journal
- IEEE Transactions on Audio, Speech and Language Processing (TASLP)
- Conferences
- ICASSP, SLT, Interspeech, ASRU, etc.
