本文提出了一种模式识别方法——改进的连续Hamming方法(ICHAM)。它从下面两个方面对使用连续资料的Hamming方法进行了改进。 1)考虑到更为一般的样品分布情况,用方差较小的那类样品计算Hamming核; 2)改进了对Hamming核的计算方法,使之处于方差较小的这一类样品的中心。 文中还给出了ICHAM法应用于北京及邻区潜在震源判定的实例。结果表明,其效果比二值Hamming方法及原来未改进的连续Hamming方法都好一些。检验表明结果较为稳定。 This paper presents a pattern recognition method - Improved Continuous Hamming Method (ICHAM). It improves on the Hamming method using continuous data from the following two aspects. 1) Hamming nuclei are calculated for samples of smaller variance, taking into account the more general distribution of the sample; and 2) the computational method for the Hamming kernel is improved so that it is at the center of this type of sample with less variance. The paper also gives an example of the ICHAM method applied to determine the potential sources in Beijing and its adjacent areas. The results show that the proposed method is better than the binary Hamming method and the original unimproved continuous Hamming method. Test results show that the more stable.