The 12-lead Electrocardiography(ECG) is the gold standard in diagnosing cardiovascular diseases, but most previous studies focused on 1-lead or 2-lead ECG. This work uses a large data set, comprising 7,704 12-lead ECG samples, as the data source, and the goal is to develop a classification model for six common types of urgent arrhythmias. We consider the characteristics of multivariate time-series data to design a novel deep learning model, combining convolutional neural network (CNN) and long short-Term memory (LSTM) to learn feature representations as well as the temporal relationship between the latent features. The experimental results indicate that the proposed model achieves promising results and outperforms the other alternatives. We also provide brief analysis about the proposed model.