摘要
为了解决金属磁记忆信号小波能量谱特征存在的相关性和冗余性问题，利用类别可分性准则，在提取金属磁记忆信号小波能量谱的基础上，将能量谱特征进行变换提取最优特征向量。将能量谱特征向量、最优特征向量和低频特征向量作为支持向量机的特征输入量分别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：最优特征向量能够减小小波能量谱特征的相关性和冗余性，有效提高支持向量机识别的准确率。
Wavelet power spectrum of metal magnetic memory signal is always correlative and re- dundant. In order to resolve this problem, the optimized eigenvector is proposed by separability theorem based on extracting wavelet power spectrum of metal magnetic memory. Then the sup- port vector machines（SVM）with wavelet power spectrum,optimized eigenvector and power spec- trum of low frequency as its input eigenvectors are introduced to recognize the metal magnetic memory signal of different areas. The result shows that the optimized eigenvector not only can e- liminate the correlation and redundancy of wavelet power spectrum,but also improve the veracity of SVM effectively.
出处
《军械工程学院学报》
2011年第6期25-28,共4页
Journal of Ordnance Engineering College
基金
军队科研计划项目
关键词
金属磁记忆
类别可分性准则
小波能量谱
特征提取
支持向量机
metal magnetic memory
separability theorem
wavelet power spectrum
feature extraction
support vector machines
