Yifei Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yifei Zhu has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
Huan Xu
Key Laboratory of Modern Acoustics and Institute of Acoustics, Nanjing University, Nanjing, 210093, China
1 shared publication- Lin Gu
Beijing National Center for Electron Microscopy and Laboratory of Advanced Materials, School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China
1 shared publication - Miao Cheng
Research Institute of Electrical Communication, Tohoku University, Sendai, Japan
1 shared publication - Mingyi Shi
The University of Hong Kong
1 shared publication - Norihiko Nishizawa
Nagoya University
1 shared publication - Shotaro Kitajima
Department of Electronics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
1 shared publication
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (6) , 1-13
DynaConTalk: Wavelet-Constrained Diffusion for Long-Form and Controllable Holistic Co-Speech 3D Motion
Holistic co-speech animation is prone to averaging in both motion representation and speech conditioning. In coordinate-space diffusion, slow body posture, mid-frequency gesture strokes, and fast hand or facial details are entangled in one prediction target, often producing low-variance, over-smoothed motion. Meanwhile, fixed multimodal conditioning can let dense rhythmic and acoustic cues dominate sparse semantic cues, weakening rare content-specific gestures. We present DynaConTalk, a wavelet-constrained diffusion framework for long-form and controllable holistic co-speech motion generation. DynaConTalk performs diffusion in stationary wavelet transform (SWT) coefficient space, where temporally aligned bands separate coarse posture evolution, gesture strokes, and high-frequency expressive details. We further introduce dynamic proposal-consensus conditioning to adaptively balance rhythmic, acoustic, semantic, textual, and identity/activity cues according to the current speech context and denoising state. Finally, to ensure smooth transitions between windows and support flexible user control, we re-noise any reference signal - whether a keypose, a motion clip, or a short boundary condition - to match the current diffusion timestep before injecting it into the denoiser. This unified strategy allows seamless long-form generation, localized editing, and style-driven guidance to be handled within a single coherent process, without task-specific modules or post-hoc stitching. Experiments show that DynaConTalk improves gesture realism, facial accuracy, temporal continuity, and controllable editing over prior holistic co-speech generation methods. Code and data for this paper are at https://zhuyifeiabcd1.github.io/DynaConTalk/.
- 9 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (6) , 1-13
Open AccessDynaConTalk: Wavelet-Constrained Diffusion for Long-Form and Controllable Holistic Co-Speech 3D Motion
Holistic co-speech animation is prone to averaging in both motion representation and speech conditioning. In coordinate-space diffusion, slow body posture, mid-frequency gesture strokes, and fast hand or facial details are entangled in one prediction target, often producing low-variance, over-smoothed motion. Meanwhile, fixed multimodal conditioning can let dense rhythmic and acoustic cues dominate sparse semantic cues, weakening rare content-specific gestures. We present DynaConTalk, a wavelet-constrained diffusion framework for long-form and controllable holistic co-speech motion generation. DynaConTalk performs diffusion in stationary wavelet transform (SWT) coefficient space, where temporally aligned bands separate coarse posture evolution, gesture strokes, and high-frequency expressive details. We further introduce dynamic proposal-consensus conditioning to adaptively balance rhythmic, acoustic, semantic, textual, and identity/activity cues according to the current speech context and denoising state. Finally, to ensure smooth transitions between windows and support flexible user control, we re-noise any reference signal - whether a keypose, a motion clip, or a short boundary condition - to match the current diffusion timestep before injecting it into the denoiser. This unified strategy allows seamless long-form generation, localized editing, and style-driven guidance to be handled within a single coherent process, without task-specific modules or post-hoc stitching. Experiments show that DynaConTalk improves gesture realism, facial accuracy, temporal continuity, and controllable editing over prior holistic co-speech generation methods. Code and data for this paper are at https://zhuyifeiabcd1.github.io/DynaConTalk/.
- Open Access
- 11 September 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (18) , 8154
A Fast Prediction Framework for Multi-Variable Nonlinear Dynamic Modeling of Fiber Pulse Propagation Using DeepONet
Traditional femtosecond laser modeling relies on the iterative solution of the Nonlinear Schrödinger Equation (NLSE) using the Split-Step Fourier Method (SSFM). However, SSFM’s high computational complexity leads to significant time consumption, particularly in automatic control and system optimization, thus limiting control model responsiveness. Recent studies have suggested using neural networks to simulate fiber dynamics, offering faster computation and lower costs. In this study, we introduce a novel fiber propagation method utilizing the DeepONet architecture for the first time. By separately managing fiber parameters and input–output pulses in the branch and trunk networks, this method can simulate various fiber configurations with high accuracy and without altering the architecture. Additionally, while SSFM generation time increases linearly with fiber length, the GPU-accelerated AI generation time remains consistent at around 0.0014 s, regardless of length. Notably, in high-order soliton (HOS) compression over a 12 m distance, the AI method is approximately 56,865 times faster than SSFM.
- 11 September 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (18) , 8154
Open AccessA Fast Prediction Framework for Multi-Variable Nonlinear Dynamic Modeling of Fiber Pulse Propagation Using DeepONet
Traditional femtosecond laser modeling relies on the iterative solution of the Nonlinear Schrödinger Equation (NLSE) using the Split-Step Fourier Method (SSFM). However, SSFM’s high computational complexity leads to significant time consumption, particularly in automatic control and system optimization, thus limiting control model responsiveness. Recent studies have suggested using neural networks to simulate fiber dynamics, offering faster computation and lower costs. In this study, we introduce a novel fiber propagation method utilizing the DeepONet architecture for the first time. By separately managing fiber parameters and input–output pulses in the branch and trunk networks, this method can simulate various fiber configurations with high accuracy and without altering the architecture. Additionally, while SSFM generation time increases linearly with fiber length, the GPU-accelerated AI generation time remains consistent at around 0.0014 s, regardless of length. Notably, in high-order soliton (HOS) compression over a 12 m distance, the AI method is approximately 56,865 times faster than SSFM.
- 11 June 2007
- Published by American Physical Society (APS) in Physical Review B
- Vol. 75 (22) , 224303
Influence of the bond defect in driven Frenkel-Kontorova chains
We study the influence of a bond defect in a nonlinear coupled pendulum chain subjected to vertical vibration experimentally. We observe that the defect can attract or repel solitons including phase-matched breather, phase-mismatched breather, and phase-mismatched kink. Furthermore, we observe a new phenomenon that the interaction polarity is dependent on the location of the defect in the phase-matched breather. If a defect locating near a breather center attracts the breather, it will repel the breather when it locates at a further place. And the effects of the defect on the phase-mismatched solitons are similar to the phenomena observed in the studies on the pendulum length defects. We also find the interaction intensity is related to the defect intensity.
- 11 June 2007
- Published by American Physical Society (APS) in Physical Review B
- Vol. 75 (22) , 224303
Influence of the bond defect in driven Frenkel-Kontorova chains
We study the influence of a bond defect in a nonlinear coupled pendulum chain subjected to vertical vibration experimentally. We observe that the defect can attract or repel solitons including phase-matched breather, phase-mismatched breather, and phase-mismatched kink. Furthermore, we observe a new phenomenon that the interaction polarity is dependent on the location of the defect in the phase-matched breather. If a defect locating near a breather center attracts the breather, it will repel the breather when it locates at a further place. And the effects of the defect on the phase-mismatched solitons are similar to the phenomena observed in the studies on the pendulum length defects. We also find the interaction intensity is related to the defect intensity.