【24h】

Computational Reduction of SPLICE using QSNR

机译:使用QSNR的拼接计算减少

获取原文

摘要

Stereo-based Piecewise Linear Compensation for Environments (SPLICE) algorithm recently developed to enhance noisy speech shows good performance in various noise environments. However, SPLICE requires great computational load to select an environment model for each frame. In this paper, without performance degradation, we propose a method to reduce the computational load on environment selection using QSNR that is a kind of local SNR. Computational load for QSNR, which involves median filtering and some comparison operations on frame energy sequence, is quietly small. We used QSNR as a feature for preselection of environments. According to experiments on the AURORA 2 database, we achieved both the drastically computational reduction and performance improvement in comparison with the conventional method.
机译:基于立体声的划分线性补偿对环境(拼接)算法最近开发的,以增强嘈杂的言论在各种噪声环境中显示出良好的性能。 但是,拼接需要很大的计算负载来为每个帧选择环境模型。 在本文中,没有性能下降,我们提出了一种使用QSNR来减少环境选择的计算负荷,这是一种本地SNR。 QSNR的计算负荷,涉及中值滤波和帧能序列上的一些比较操作,是悄然的小。 我们使用QSNR作为预选环境的功能。 根据Aurora 2数据库的实验,我们与传统方法相比,我们达到了急剧计算的减少和性能改进。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号