Nonlinear real-life signal detection with a supervised principal components analysis.
Nonlinear real-life signal detection with a supervised principal components analysis.
复制标题
DOI：
10.1063/1.2437579
复制
发表时间：
2007-02
期刊：
影响因子：
2.9
通讯作者：
C. T. Zhou;T. X. Cai;T. Cai
中科院分区：
文献类型：
--
作者：
C. T. Zhou;T. X. Cai;T. Cai
研究了一种新的检测方法--监督主成分分析法，用于检测未知噪声环境中的目标信号。在我们的检测方案中有两个通道。每个通道分别由非线性相空间重建器（用于使用接收到的时间序列嵌入数据矩阵）和主成分分析器（用于特征提取）组成。输出误差的时间序列，这是由于这两个通道的相关数据矩阵的两个特征向量的差异，然后使用时频工具，例如，频谱或Wigner-Ville分布进行分析。基于实际电磁数据的实验结果验证了该算法的检测性能。发现可以检测到隐藏在本底噪声下的微弱信号。此外，检测性能的鲁棒性清楚地表明，当信号功率不太低时，可以提取信号频率。
A novel strategy named supervised principal components analysis for the detection of a target signal of interest embedded in an unknown noisy environment has been investigated. There are two channels in our detection scheme. Each channel consists of a nonlinear phase-space reconstructor (for embedding a data matrix using the received time series) and a principal components analyzer (for feature extraction), respectively. The output error time series, which results from the difference of both eigenvectors of the correlation data matrices from these two channels, is then analyzed using time-frequency tools, for example, frequency spectrum or Wigner-Ville distribution. Experimental results based on real-life electromagnetic data are presented to demonstrate the detection performance of our algorithm. It is found that weak signals hidden beneath the noise floor can be detected. Furthermore, the robustness of the detection performance clearly illustrated that signal frequencies can be extracted when the signal power is not too low.