Federated Learning with Feature Space Separation for Speaker Recognition

FedFSS

Training

python train_fed_anchor_lama_gamma.py

Environment

torch 1.12.1; torchaudio 0.12.1; torchvision 0.13.1
pip install soundfile, tqdm, matplotlib, Pillow, scikit-learn, numpy

Result

Training on Vox1

Method	Client1	Client2	Client3	Client4
baseline	6.45	6.79	6.29	6.99
Fedavg	4.86	4.94	4.98	5.62
FedFSS	4.72	4.65	4.57	5.07

Citation

@inproceedings{meng25_interspeech,
  title     = {{Federated Learning with Feature Space Separation for Speaker Recognition}},
  author    = {Ying Meng and Zhihua Fang and Liang He},
  year      = {2025},
  booktitle = {{Interspeech 2025}},
  pages     = {1518--1522},
  doi       = {10.21437/Interspeech.2025-364},
  issn      = {2958-1796},
}

I'm so glad that our sharing was useful to you.