Disentangling Voice and Content with Self-Supervision for Speaker Recognition

Authors: TIANCHI LIU, Kong Aik Lee, Qiongqiong Wang, Haizhou Li

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The efficacy of the proposed framework is validated via experiments conducted on the Vox Celeb and SITW datasets with 9.56% and 8.24% average reductions in EER and min DCF, respectively.
Researcher Affiliation Academia 1 Institute for Infocomm Research (I2R), Agency for Science, Technology and Research (A STAR), Singapore 2 Dept. of Electrical and Computer Engineering, National University of Singapore, Singapore 3 Dept. of Electrical and Electronic Engineering, Hong Kong Polytechnic University, Hong Kong 4 School of Data Science, The Chinese University of Hong Kong, Shenzhen, China
Pseudocode No No explicitly labeled pseudocode or algorithm blocks found.
Open Source Code No The paper does not contain any statement or link providing concrete access to the source code for the described methodology.
Open Datasets Yes The experiments are conducted on Vox Celeb1 [54], Vox Celeb2 [13], and the Speaker in the Wild (SITW) [48] datasets.
Dataset Splits Yes The experiments are conducted on Vox Celeb1 [54], Vox Celeb2 [13], and the Speaker in the Wild (SITW) [48] datasets. ... The development set for training is Vox Celeb2 Dev Set and we use Vox Celeb1 test set for evaluation.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts) are provided for the experimental setup.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, library versions) are explicitly mentioned in the paper.
Experiment Setup Yes All the models are evaluated by the performance in terms of equal error rate (EER) and the minimum detection cost function (min DCF). Detailed descriptions of datasets, training strategy, and evaluation protocol are available in Appendix C. ... We train our models with an Adam optimizer [30] with a learning rate of 0.001.