CV

Ph.D.

National University of Singapore

Computer Science

M.Eng.

National University of Singapore

Mechanical Engineering

B.Eng. / B.A.

Shanghai Jiao Tong University

Energy & Power Engineering (Major); Musicology (Minor)

DuET: Note-Frame Dual Encoding for Piano Performance-to-Score Transcription

arXiv preprint

Note-frame dual-encoder Transformer jointly learns PM2S transcription and beat/downbeat tracking via bidirectional cross-attention; TimeRoPE and TPRoPE encode relative time and pitch, reducing MUSTER error from 6.56% to 3.64% on the curated Gangqinpu dataset.

Bridging Piano Transcription and Rendering via Disentangled Score Content and Style

ICLR 2026

Unified transformer Seq2Seq jointly models APT and EPR by disentangling score content and performance style; diffusion-based style recommender enables alignment-free training and style transfer.

CoMelSinger: Discrete Token-Based Zero-Shot Singing Synthesis With Structured Melody Control and Guidance

IEEE TASLP

MaskGCT-based zero-shot singing synthesis with disentangled melody control; coarse-to-fine contrastive learning limits prosody leakage and an encoder-only SVT module supplies frame-level pitch supervision.

Peransformer: Improving Low-informed Expressive Performance Rendering with Score-aware Discriminator

APSIPA ASC 2025

Score-aware discriminator and transformer framework enhance expressive performance rendering when score information is sparse.

Automatic Lyric Transcription and Automatic Music Transcription from Multimodal Singing

ACM TOMM

Multimodal ALT/AMT with audio and lip video using residual cross-attention; first N20EM dataset and SSL encoders yield noise-robust lyric and note transcription.

End-to-End Real-World Polyphonic Piano Audio-to-Score Transcription with Hierarchical Decoding

IJCAI 2024

Hierarchical decoder predicts bar-level key/time and note content; two-stage training with EPR pretraining then human-performance finetuning bridges synthetic–real gap.

Elucidate Gender Fairness in Singing Voice Transcription

MM 2023

Reveals gender bias in singing voice transcription from pitch distributions; adversarial attribute predictor with note-conditioned alignment cuts bias >50% with minimal accuracy loss.

Dual-Encoder Piano Performance-to-Score Transcription

Proposed DuET, a note-frame dual-encoder Transformer jointly learning note-level PM2S transcription and frame-level beat/downbeat tracking; introduced TimeRoPE and TPRoPE; curated the 18.8K-pair Gangqinpu popular piano dataset and released an audio-to-score demo.

Joint Piano Transcription and Performance Rendering

Unified transformer Seq2Seq for APT+EPR; alignment-free formulation and diffusion style recommendation.

Hierarchical Audio-to-Score Transcription for Piano

End-to-end Seq2Seq with hierarchical decoder and two-stage training using EPR pretraining then real-perf finetuning.

Tencent Singapore

Speech Synthesis Intern (Mentor: Dr. Chenglin Xu)

Integrating singing voice synthesis into a spoken dialogue system, enabling it to sing on request within natural conversational interactions.

ByteDance Singapore (TikTok E-commerce)

Algorithm Engineer Intern, Supply Chain and Logistics (Mentor: Dr. Chengxin Wang)

Developed ML/OR models for GMV and inventory forecasting and built ETA/logistics-event prediction and NLP-based address-processing pipelines to support demand planning and fulfillment for TikTok Shop Europe.

SoC Honour List of Student Tutors

Teaching Fellowship of SoC

Research Achievement Award of SoC

President's Graduate Fellowship

Academic Excellence Scholarship of SJTU