【24h】

Blind Separation of Sources from Their Delayed Mixtures

机译:从延迟混合物中盲分离源

获取原文
获取原文并翻译 | 示例

摘要

We give a near real-time algorithm for the blind separation of sources from mixtures of delayed sources. Many successful algorithms have been proposed for blind separation of instantaneous mixtures. For delayed mixtures only a few algorithms have been reported, and they assume only cross delays between mixed sources. Moreover, the weights and cross delays are updated in a batch mode, which is not suitable for real-time operation. The algorithm given here allows for delays of all the sources in the mixtures, which is necessary for most practical blind source separation problems. A near real-time procedure to update delays and weights is also proposed. Furthermore, in the presence of sensor noise, previously reported algorithms converge to incorrect cross delay values. This paper gives a wavelet based denoising method that significantly improves convergence to correct delays, thereby improving convergence to correct delays and weights.
机译:我们给出了一种近实时算法,用于从延迟源的混合物中盲目分离源。已经提出了许多成功的算法,用于瞬时混合物的盲分离。对于延迟混合,仅报告了少数算法,并且它们仅假设混合源之间的交叉延迟。此外,权重和交叉延迟以批处理模式进行更新,这不适用于实时操作。此处给出的算法允许混合物中所有源的延迟,这对于大多数实际的盲源分离问题是必需的。还提出了一种更新延迟和权重的近实时程序。此外,在存在传感器噪声的情况下,先前报告的算法收敛到不正确的交叉延迟值。本文提出了一种基于小波的去噪方法,该方法可显着改善收敛以纠正延迟,从而改善收敛以纠正延迟和权重。

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