...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Signal Processing: The Official Publication of the European Association for Signal Processing (EURASIP) >BLIND SOURCE SEPARATION FOR CONVOLUTIVE MIXTURES
【24h】

BLIND SOURCE SEPARATION FOR CONVOLUTIVE MIXTURES

机译：BLIND SOURCE SEPARATION FOR CONVOLUTIVE MIXTURES

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相关主题

摘要

Blind separation of sources is now a well known problem, and various methods have been proposed for instantaneous mixtures, and convolutive mixtures of narrow-band sources. In this paper, we present adaptive algorithms for separation of wide-band signals, for convolutive mixtures modeled by Finite Impulse Response (FIR) filters. The algorithms estimate coefficients of FIR filters using independence criteria. Under the condition of fourth-order white noise, we prove in this paper that the estimation of filter coefficients may be based on the cancellation of 4th-order output cross-cumulants. However, simpler algorithms only using a few fourth-order cross-cumulants or moments, and 2nd-order moments have the same efficacy. Finally, experimental results point out good performance(about -20 dB of residual crosstalk) even with nonwhite and nonstationary signals like speech signal for simulated mixtures. For recorded mixtures, results are not as good and suggest that the FIR model of mixtures is not realistic enough and must be improved.

著录项

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