磁记忆信号处理和特征量提取是金属磁记忆检测的核心。传统方法采用寻找磁记忆信号法向分量过零点区域，并且用梯度值大小衡量应力集中程度。对带有焊接缺陷的10号钢管件进行了磁记忆检测试验，对Lipschitz指数的机理进行了分析，并利用该方法对磁记忆检测信号进行了特征分析。结果表明，单纯用过零点和梯度值不能完全反映磁记忆信号特征。在局部区域，凡是小于设定的奇异性指数，均可判断存在故障；根据信号的奇异性与模极大值存在的关系，通过对小波变换系数模极大值的判断，就可以找到信号突变位置。

Signal processing and characteristic quantity picking-up are the kernel of the Metal Magnetic Memory Technology. Traditional method of measuring the degree of stress concentration is finding areas where vertical quantity of detection signal of the metal magnetic memory goes through zeros and measuring the degree of stress concentration with gradient of vertical quantity of detection signal. Magnetic memory testing on the steel tube with weld faults was carried out, and the mechanism of Lipschitz was analyzed. Results indicated that areas where vertical quantity passed through zeros and gradient of detection signal of the metal magnetic memory could not reflect the entirety of detection signal. In local areas, all areas whose singularity index was smaller than singularity index were indications of faults. It was concluded that one could find positions of signal singularity through judging module maximum of wavelet transform coefficients in view of the relation between singularity and module maximum.<收到修改稿日期2008-04-28>