I am a Research Scientist at Shanda Group Corp, working on speech and language AI.

My research studies reliable speech intelligence in complex acoustic scenes, spanning robust speech processing, multi-talker speech understanding, and speech language models. I am particularly interested in how speech models can exploit increasingly powerful learned priors while remaining faithful to the underlying acoustic evidence.

• Reliable Perception: Robust Speech Processing
  Recover reliable speech evidence from noisy, reverberant, and otherwise degraded observations.
  Speech Enhancement · Noise-Robust ASR · Dereverberation

• Evidence Attribution: Multi-Talker & Speaker-Grounded Speech
  Understand what was said, who said it, and which acoustic evidence belongs to which source in overlapping speech scenes.
  Multi-Talker ASR · Speaker Grounding · Separation / TSE

• Grounded Modeling: Speech LLMs & Spoken Interaction
  Leverage language-model and generative priors while keeping model decisions grounded in the observed speech.
  Speech LLMs · Grounded Generation · Speech-to-Speech

Multi-Talker ASR

Serialized Output Prompting for Large Language Model-based Multi-Talker Speech Recognition

IEEE ASRU, 2025

Proposes serialized output prompting to improve large language model-based recognition in overlapping multi-talker speech.

Paper | BibTeX

Speech Enhancement

Combining Deterministic Enhanced Conditions with Dual-Streaming Encoding for Diffusion-Based Speech Enhancement

IEEE/ACM TASLP, 2025

Introduces a diffusion-based speech enhancement framework with deterministic enhanced conditions and dual-stream encoding.

Paper | BibTeX

Robust Speech Processing

Waveform-domain Speech Enhancement Using Spectrogram Encoding for Robust Speech Recognition

IEEE/ACM TASLP, 2024

Presents a waveform-domain speech enhancement method with spectrogram encoding for robust speech recognition.

Paper | BibTeX