In response to the poor performance of traditional electronic information signal denoising techniques in low signal-to-noise ratio (SNR) environments, this study introduces a deep learning based denoising method aimed at improving the accuracy and robustness of signal processing in complex environments. Firstly, this study designs and implements a deep convolutional neural network (CNN) model specifically for identifying and preprocessing Gaussian white noise in electronic information signals. This model extracts the features of the original signal from multiple convolutional layers, and the pooling layer gradually reduces the size of the feature map, thereby reducing noise interference and preserving key information. The feature map output by the convolutional layer is processed by the non-linear activation function ReLU (Rectified Linear Unit), which introduces non-linear factors and avoids the limitation of linear relationships in the feature map on the model’s expressive ability. Then, in order to further extract effective features from the denoised signal and suppress residual noise, a structure based on stacked autoencoders is adopted, which are trained to minimize the reconstruction error between the input and output. The study compares the denoising effect of radar and communication signals under the same noise, and finds that the SNR of radar signals is significantly improved by 6 dB after model processing, which is better than the 4 dB of communication signals. By introducing deep learning techniques, this study has successfully developed a new method for denoising electronic information signals, which not only improves denoising efficiency but also expands the applicability of the method.