An automatic identification method of foreign object sound in electric energy meter based on variable step size least mean square（ LMS） and support vector machine（ SVM） is proposed. In consideration of the fact that the SVM classifier is sensitive to noise

noise reduction of the foreign object sound signal in the collected electric energy meter is realized by using the variable step size LMS. Compared with the fixed step length LMS

the signal to noise ratio is improved significantly and the time consumption is reduced. The sound signal is analyzed in time

frequency and cepstrum domain

and its short-term characteristic coefficient and improved Meier frequency cepstrum coefficient（ MFCC） coefficient is extracted. The short-time energy and MFCC coefficients are used to form mixed feature matrix

which is input a SVM for foreign object sound recognition after dimension reduction. The experimental results indicate that the method has small computational amount

high recognition rate

and has good application value.