摘要
提出一种针对周期性源信号的单通道盲源分离方法,该方法首先利用希尔伯特变换将单通道的混合信号表示为复数形式后,并用交叉互验技术来估计谐波分量的阶数,然后通过人工蜂群算法优化用于估计基频的代价函数以获得源数、基频及其谐波长度,最后由一种自适应滤波算法来估算源信号的幅值,从而实现了盲源分离的目的.仿真实验结果表明了该方法思路的可行性,且可达到"全盲"的要求.
A method of single-channel blind separation for periodic source signals was proposed. Firstly, the mixed signal was represent as complex form, and an algorithm based on cross validation technology to estimate number of harmonic components was introduced. A goal function of fundamental frequency was optimized by using artificial bee colony optimization, and then it led to determine number of sources, fundamental frequency and length of harmonic components of each source. At last, amplitude of sources was estimated using an algorithm of adaptive filter. The source signals were separated from mixed single-channel through above programs. The simulation results showed that the method can achieve the single-channel blind separation for source signals, and it met the requirement of "full-blind".
出处
《集美大学学报(自然科学版)》
CAS
2014年第1期75-80,共6页
Journal of Jimei University:Natural Science
基金
国家自然科学基金资助项目(51309116
51179074)
集美大学科研基金资助项目(ZQ2013001
ZC2013012)
人工智能四川省重点实验室开放基金项目(2014RYJ03)
关键词
单通道
盲源分离
交叉互验
人工蜂群优化算法
周期信号
希尔伯特变换
single-channal
blind source separation
cross validation
artificial bee colony optimization
periodic sources
Hilbert transform