Traditional acoustic models have limitations such as complex components, difficulty in joint training, and the need for data pre-alignment. To this end, this paper proposes an end-to-end Chinese speech recognition model that combines a 1D gated convolutional neural network with Connectionist Temporal Classification(CTC). the core of the model consists of stacked multi-layer 1D convolutional networks to extract contextual high-level features, gated linear units (GLU) to suppress gradient dispersion, and CTC to achieve end-to-end training and decoding of Chinese characters at the character level. Experiments demonstrate that the model significantly improves the performance over the baseline model on the public dataset, with a CER reduction above of 2.5%.