9 papers
FlowGRPO for Vocal Separation: From Reward Analysis to Reward Design
Lishi Zuo, Chong-Xin Gan, Zezhong Jin +3
The Flow Matching objective supervises the velocity field at intermediate states, whereas waveform metrics are evaluated after decoding and may be non-differentiable. To align Flow…
Decoupled Latent Flow Matching for Few-Step Joint Vocal-Accompaniment Separation
Lishi Zuo, Youzhi Tu, Lu Yi +4
Generative modeling provides a flexible way to model mixture-conditioned source distributions, but iterative diffusion and flow matching models are costly for long music signals. T…
Beyond Residual Connections: Manifold-Constrained Hyper-Connections for Robust Speaker Representation Learning
Zezhong Jin, Xiaoyu Wang, Zhe Li +4
Residual connections are fundamental to deep speaker recogni- tion models, such as ECAPA-TDNN and ResNet. However, standard identity mapping limits information flow to a sin- gle p…
EII-SCL: Harnessing Emotional Inertia for Multimodal Emotion Recognition in Conversation
Zilong Huang, Kong Aik Lee, Chong-Xin Gan +3
Multimodal emotion recognition in conversation (MERC) achieves accurate predictions by integrating multimodal and contextual information in dialogues. While current MERC approaches…
LISE : Listenable Interpretable Speaker Embeddings
Xiaoliang Wu, Chongxin Gan, Ke Liu +2
Deep neural network-based automatic speaker verification (ASV) systems achieve impressive performance but their embedding representations remain opaque, lacking a structured and pe…
Heterogeneity-Aware Dataset Scheduling for Efficient Audio Large Language Model Training
Yanru Wu, Jianning Wang, Chongxin Gan +1
Training general-purpose Audio Large Language Models (ALLMs) across diverse datasets is essential for holistic audio understanding, yet it faces significant challenges due to datas…
UNet-Based Fusion and Exponential Moving Average Adaptation for Noise-Robust Speaker Recognition
Chong-Xin Gan, Peter Bell, Man-Wai Mak +4
The joint training of speech enhancement and speaker embedding networks for speaker recognition is widely adopted under noisy acoustic environments. While effective, this paradigm…
TrInk: Ink Generation with Transformer Network
Zezhong Jin, Shubhang Desai, Xu Chen +8
In this paper, we propose TrInk, a Transformer-based model for ink generation, which effectively captures global dependencies. To better facilitate the alignment between the input…
Asymmetric Clean Segments-Guided Self-Supervised Learning for Robust Speaker Verification
Chong-Xin Gan, Man-Wai Mak, Weiwei Lin +1
Contrastive self-supervised learning (CSL) for speaker verification (SV) has drawn increasing interest recently due to its ability to exploit unlabeled data. Performing data augmen…