In this paper, the data set for human activity recognition collected by the wearable device is processed, and the processed data is used to train a one-dimensional convolutional neural network to test and obtain accuracy results. The processing of the data set allows some noise and invalid data in the original data set to be filtered out, which reduces the amount of calculation and improves the efficiency of the neural network when training the neural network. After testing, under the condition that the structure of the neural network is unchanged, the processed data set can improve the performance of the neural network.