Dual Data Scaling for Robust Two-Stage User-Defined Keyword Spotting
Scaling acoustic data and phoneme-matching anchors for robust user-defined wake words.
Publications
Publications are grouped by research area, with current manuscripts identified separately. See also Google Scholar.
Speech & Audio Intelligence
Keyword Spotting & Wake Words
Effective User-Defined Keyword Spotting With Dual-Stage Matching, Multi-Modal Enrollment, and Continual Adaptation
Dual-stage matching, multimodal enrollment, and lightweight adaptation for personalized keyword spotting.
MM-KWS: Multi-modal Prompts for Multilingual User-defined Keyword Spotting
Text and speech enrollment for multilingual, user-defined keyword spotting.
Speaker Recognition & Modeling
Stabilizing Short Duration Speaker Verification through Neural Re-scoring with Hybrid Enrollment
Hybrid enrollment and neural re-scoring for reliable verification from short speech segments.
Open-Set Speaker Identification Through Efficient Few-Shot Tuning With Speaker Reciprocal Points and Unknown Samples
Few-shot open-set speaker identification with reciprocal points and synthesized or real unknown samples.
VoxAging: Continuously Tracking Speaker Aging with a Large-Scale Longitudinal Dataset in English and Mandarin
A longitudinal dataset for studying speaker aging and its effects on speaker verification.
Towards Robust Speaker Recognition against Intrinsic Variation with Foundation Model Few-shot Tuning and Effective Speech Synthesis
Few-shot tuning and speech synthesis to address aging and emotional variation in speaker recognition.
Enhancing Open-Set Speaker Identification Through Rapid Tuning With Speaker Reciprocal Points and Negative Sample
Rapid enrollment-time tuning for open-set speaker identification.
Text-to-Speech Synthesis
StableTTS: Towards Efficient Denoising Acoustic Decoder for Text to Speech Synthesis with Consistency Flow Matching
Consistency flow matching for efficient denoising acoustic decoding in text-to-speech synthesis.
StyleFusion TTS: Multimodal Style-Control and Enhanced Feature Fusion for Zero-Shot Text-to-Speech Synthesis
Multimodal style and speaker control with enhanced feature fusion for zero-shot speech synthesis.
Optimizing feature fusion for improved zero-shot adaptation in text-to-speech synthesis
* Equal contribution.
A two-branch speaker control module for improved voice cloning and zero-shot speaker adaptation.
Speech Bandwidth Extension
Bridge, Distill, Refine: High-Fidelity One-Step Generation for Multi-Bandwidth Speech Extension
One-step, multi-bandwidth speech extension through a cutoff-aware flow bridge, progressive consistency distillation, and adversarial refinement.
Speech Language Models
PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs
Phoneme-level retrieval and selective correction for contextual biasing in speech language models.
Embodied Audio
Acoustic Simulation
AcousticAgent: A Floorplan-Centric Simulation Platform for Distributed Multi-Room Speech Processing
Floorplan-centric acoustic simulation for editable, connected multi-room scenes and distributed speech processing, with the FP-RIR dataset spanning 15,376 residential floorplans.
Interdisciplinary Collaborations
Wireless Positioning
Enhanced Fingerprint-Based Positioning With Practical Imperfections: Deep Learning-Based Approaches
Collaborative work on deep learning for wireless fingerprint positioning under practical imperfections.
Computer Vision
StereoDETR: Stereo-Based Transformer for 3D Object Detection
Collaborative work on efficient stereo-based 3D object detection.
DDStereo: Efficient Dual Decoder Transformers for Stereo 3D Road Anomaly Detection
Dual-decoder stereo transformers for efficient 3D object detection and out-of-distribution road anomaly detection.
Stage-wise Adaptive Label Distribution for Facial Age Estimation
Stage-wise adaptive label distributions for modeling ambiguity in facial age estimation.