we use the public dataset 2008 BCI competition IV-2a to analyze the training and evaluation datasets of nine objects. We first obtain the event from each session, generate epochs, and mark the energy set for different time periods. All the pre-processing data are generated according to the unified energy scale and six types of custom energy calculation methods. Next we generated the training topographical maps and input them into the CNN network for training. Then, a classifier supporting multi-object classification is achieved. Finally, topographical maps of evaluation datasets were judged, and the classification accuracy is obtained statistically.