期刊
IEEE SIGNAL PROCESSING LETTERS
卷 28, 期 -, 页码 1265-1269出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LSP.2021.3089437
关键词
Feature extraction; Speech synthesis; Training; Mel frequency cepstral coefficient; Task analysis; Standards; Neural networks; Synthetic speech detection; speech forensics; ASVspoof2019; ASVspoof2015; cross-dataset testing; end-to-end
资金
- 2020-2021 International Scholar Exchange Fellowship (ISEF) Program at theChey Institute forAdvanced Studies, South Korea
- National Natural Science Foundation of China [61802284]
This paper introduces a new synthetic speech detection approach, the TSSDNet model, which utilizes end-to-end DNN and eliminates the need for hand-crafted feature extraction. Experimental results show significant performance improvement, demonstrating the potential and advantages of this model in synthetic speech detection.
The constant Q transform (CQT) has been shown to be one of the most effective speech signal pre-transforms to facilitate synthetic speech detection, followed by either hand-crafted (subband) constant Q cepstral coefficient (CQCC) feature extraction and a back-end binary classifier, or a deep neural network (DNN) directly for further feature extraction and classification. Despite the rich literature on such a pipeline, we show in this paper that the pre-transform and hand-crafted features could simply be replaced by end-to-end DNNs. Specifically, we experimentally verify that by only using standard components, a light-weight neural network could outperform the state-of-the-art methods for the ASVspoof2019 challenge. The proposed model is termed Time-domain Synthetic Speech Detection Net (TSSDNet), having ResNet- or Inception-style structures. We further demonstrate that the proposed models also have attractive generalization capability. Trained on ASVspoof2019, they could achieve promising detection performance when tested on disjoint ASVspoof2015, significantly better than the existing cross-dataset results. This paper reveals the great potential of end-to-end DNNs for synthetic speech detection, without hand-crafted features.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据