Wearable ECG monitoring devices have been used in clinical applications, but the hardware configuration of wearable ECG monitoring devices limits the application of large deep learning models, which hinders the development of their intelligence. Research on the use of small convolutional neural networks for the detection of ECG signals is of interest for the development of wearable ECG devices as well as smart medicine. In this study, a one-dimensional small convolutional neural network (1D-SCNN) including five convolutional layers, four max pooling layers, a two-layer pooling module, and two fully connected layers is proposed. This study implements automatic detection of four types of arrhythmias and discusses the effects of single-path multilayer networks (depth) and multi-path networks (width) on the performance of the small convolutional neural network.