Paper ID | SPE-37.4 | ||
Paper Title | A JOINT TRAINING FRAMEWORK OF MULTI-LOOK SEPARATOR AND SPEAKER EMBEDDING EXTRACTOR FOR OVERLAPPED SPEECH | ||
Authors | Naijun Zheng, The Chinese University of Hong Kong, Hong Kong SAR China; Na Li, Bo Wu, Meng Yu, Tencent AI Lab, China; JianWei Yu, The Chinese University of Hong Kong, Hong Kong SAR China; Chao Weng, Dan Su, Tencent AI Lab, China; XunYing Liu, Helen Meng, The Chinese University of Hong Kong, Hong Kong SAR China | ||
Session | SPE-37: Speaker Recognition 5: Neural Embedding | ||
Location | Gather.Town | ||
Session Time: | Thursday, 10 June, 14:00 - 14:45 | ||
Presentation Time: | Thursday, 10 June, 14:00 - 14:45 | ||
Presentation | Poster | ||
Topic | Speech Processing: [SPE-SPKR] Speaker Recognition and Characterization | ||
IEEE Xplore Open Preview | Click here to view in IEEE Xplore | ||
Abstract | In multi-talker cases, overlapped speech degrades the speaker verification (SV) performance dramatically. To tackle this challenging problem, speech separation with multi-channel techniques can be adopted to extract each speaker's signals to improve the SV performance. In this paper, a joint training framework of the front-end multi-look speech separator and the back-end speaker embedding extractor is proposed for multi-channel overlapped speech. To better leverage the complementarity between the speech separator and the speaker embedding extractor, several training strategies are proposed to jointly optimize the two modules. Experimental results show that the proposed joint training framework significantly outperforms the individual SV system by around 52% relative EER reduction. Additionally, the robustness of the proposed framework is further evaluated under different conditions. |