摘要
介绍了一种混沌时间序列的非线性动力学降噪方法——变尺度概率净化法,并将之优化应用于混响背景下目标回波的提取。该方法以近似条件下不含目标回波的混响数据作为参考,使目标回波分离后的混响背景数据与参考混响数据在相空间上具有一致的概率分布,从而实现目标回波的提取。在信混比不低于6dB的情况下,该方法具有较好的目标回波提取效果。文中对算法作了一些改进,降低了算法的运算量和运行所需计算机内存,而降噪性能并没有下降。
A nonlinear noise reduction method for chaotic time series, the scaled probabilistic cleaning method, is introduced. It is optimized and applied to the extraction of target echoes form reverberation. Let reverberation under the same condition but in the absence of target echoes be a reference. The probabilistie distribution of reverberation data obtained by extracting the echo from the actual signal is made identical to that of the reference. When the signal-to-reverberation ratio is more than 6dB, this method works very well. This paper proposes a modification to the method, which reduces the computation complexity and the required memory size without sacrificing the noise reduction performance.
出处
《声学技术》
CSCD
北大核心
2006年第3期192-196,共5页
Technical Acoustics
基金
国家基础研究项目(5132102ZZT32)
国家重点实验室基金项目(514450801JB1101)资助课题
关键词
非线性动力学
降噪
信号分离
nonlinear dynamics
noise reduction
signal separation