THUEE system description for NIST 2020 SRE CTS challenge
This is an incremental improvement for speaker recognition systems in telephony applications.
The paper tackles speaker recognition on conversational telephone speech by developing subsystems with ResNet74, ResNet152, and RepVGG-B2 as embedding extractors, using a CM-Softmax loss and two-staged training, resulting in a first-place ranking in the NIST 2020 SRE CTS challenge.
This paper presents the system description of the THUEE team for the NIST 2020 Speaker Recognition Evaluation (SRE) conversational telephone speech (CTS) challenge. The subsystems including ResNet74, ResNet152, and RepVGG-B2 are developed as speaker embedding extractors in this evaluation. We used combined AM-Softmax and AAM-Softmax based loss functions, namely CM-Softmax. We adopted a two-staged training strategy to further improve system performance. We fused all individual systems as our final submission. Our approach leads to excellent performance and ranks 1st in the challenge.