基于Wigner-Ville分布及局部奇异值分解的磁记忆信号特征提取与识别

针对磁记忆检测中缺陷信号持续时间短且频率范围小的特点，为提取磁记忆信号的有效特征，根据矩阵奇异值的特点，提出一种基于Wigner-Ville分布及局部奇异值分解的磁记忆信号特征提取方法．通过将时频分布矩阵从时间轴和频率轴分别划分为不同局部矩阵，提取出各矩阵的奇异值来构造特征向量．然后，将构造的特征向量作为支持向量机的输入向量对不同检测区域的金属磁记忆信号进行识别．实验结果表明：基于Wigner—Ville分布及局部奇异值分解算法构造的特征向量能有效提取磁记忆信号的特征信息，提高支持向量机的识别精度．

Feature Extraction and Recognition of Metal Magnetic Memory Signal Based on Wigner-Ville Distribution and Partial Singular Values Decomposition

A new algorithm for the effective characteristics extraction of magnetic memory signal is proposed with consideration of its short duration and narrow frequency range. The proposed method is based on Wigner-Ville distribution and partial singular values decomposition. The eig- envectors are constructed by extracting singular values of several partial matrixes which are ob- tained through dividing the time and frequency axes of time-frequency matrix into different parts. Then the support vector machine （SVM） with the constructed eigenvectors as its input eigenvee- tots is employed to recognize the metal magnetic memory signals of different areas. The experi- mental result indicates that the characteristics of magnetic memory signal are efficiently extracted and the recognition precision of SVM is also improved.