Clinical examination of ectopic beats is very important for early detection, diagnosis and treatment of cardiovascular diseases. Automatic recognition of ectopic beats can effectively reduce the burden of manual recognition. In this paper, the 10 minutes Electrocardiogram(ECG) signals of 37 mice were used for analysis. All ectopic beats were calibrated by 3 experts to establish the database. Using 7 machine learning methods, the ectopic beats were automatically identified by combining the values of Impulse Rejection Filter(IRF) and the template matching algorithm. The experimental results show that 7 machine learning methods can achieve good predictive performances(all Area Under Curve(AUC)>0.899), where the ensemble learning method－AdaBoost has the best predictive performance(AUC=0.940,sensitivity=specificity=0.888).