A Supervised Locality Preserving Projection (SLPP) method is employed for channel compensation in an i-vector based speaker verification system. SLPP preserves more important local information by weighing both the within- and between-speaker nearby data pairs based on the similarity matrices. In this paper, we propose an improved SLPP (P-SLPP) to enhance the channel compensation ability. First, the conventional Euclidean distance in conventional SLPP is replaced with Probabilistic Linear Discriminant Analysis (PLDA) scores. Furthermore, the weight matrices of P-SLPP are generated by using the relative PLDA scores of within- and between-speaker pairs. Experiments are carried out on the five common conditions of NIST 2012 speaker recognition evaluation (SRE) core sets. The results show that SLPP and the proposed P-SLPP outperform all other state-of-the-art channel compensation methods. Among these methods, P-SLPP achieves the best performance.
Cite as: You, L., Guo, W., Song, Y., Zhang, S. (2018) Improved Supervised Locality Preserving Projection for I-vector Based Speaker Verification. Proc. Interspeech 2018, 62-66, doi: 10.21437/Interspeech.2018-41
@inproceedings{you18_interspeech,
author={Lanhua You and Wu Guo and Yan Song and Sheng Zhang},
title={{Improved Supervised Locality Preserving Projection for I-vector Based Speaker Verification}},
year=2018,
booktitle={Proc. Interspeech 2018},
pages={62--66},
doi={10.21437/Interspeech.2018-41}
}