In this paper, a sound source localization model based on residual network and channel attention module is proposed.In this method, the residual structure and channel attention module are combined with the convolution layer of CRNN. The short-time Fourier phase spectrum is used as the input feature, and the advanced features are extracted and filtered by the residual structure and channel attention module, so as to obtain better localization performance.To illustrate the reliability of the proposed model, we first compare it with the popular Convolutional Recurrent Neural Network (CRNN)-based localization framework on a publicly available dataset, where our proposed model shows better performance in terms of localization accuracy and error, and also set up a comparison model to verify the effect of the channel attention module on the localization effect.On the other hand, we explored the performance of the model in the face of real data using microphone array signals collected in a real environment, and compared and analyzed the accuracy under seven different features, and the results showed that the combination of log-Mel spectrum and GCC-PHAT features possessed a more obvious advantage over the rest of the features, and the amplitude spectrum and log-Mel spectrum were the least effective in real data.