https://github.com/slinusc/speaker_recognition_evaluation
Investigating Layer-Specific Performance in Speaker Recognition with XLS-R Architecture
https://github.com/slinusc/speaker_recognition_evaluation
hidden-states speaker-recognition xls-r
Last synced: 4 months ago
JSON representation
Investigating Layer-Specific Performance in Speaker Recognition with XLS-R Architecture
- Host: GitHub
- URL: https://github.com/slinusc/speaker_recognition_evaluation
- Owner: slinusc
- Created: 2024-04-15T12:45:00.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-04-29T17:46:27.000Z (about 1 year ago)
- Last Synced: 2025-01-06T08:31:31.228Z (5 months ago)
- Topics: hidden-states, speaker-recognition, xls-r
- Language: Jupyter Notebook
- Homepage:
- Size: 1.6 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
### Abstract
This study explores the impact of various layers of the XLS-R encoder on speaker recognition accuracy using CNNs and logistic regression. Findings indicate that earlier layers yield higher accuracy, highlighting their importance in feature capture. The study also reveals a significant gender disparity in accuracy. These results suggest the need for further investigation into model biases and optimizations.