Research on interference suppression algorithm of UAV relay communication based on machine learning

In this paper, a novel interference suppression algorithm for UAVS relay communication based on machine learning is proposed to improve the anti-interference ability of UAVS communication system by using Welch spectrum and convolutional neural network (CNN). The algorithm first normalizes the input signal, then calculates the Welch spectral sequence and converts it into decibel values to enhance the contrast between power values. As a good input object of CNN, Welch spectrum sequence not only has less computation, but also performs well in classification effect. A deep neural network (DNN) classifier is used to classify and recognize the extracted features, and finally the signal category is determined. In addition, this paper also introduces a contrast learning method based on data enhancement, which enhances the semantic information of the network by ensuring that the features of the same class samples are relatively close in the embedded space. Specifically, the K-nearest neighbor algorithm is used to select positive and negative sample pairs, and the batch gradient descent method is used for contrast learning in the training process to strengthen the boundaries between classes. In addition, through the double contrast mechanism in feature and cluster dimensions, the feature differentiation ability and the classification ability of the model for complex signals are further improved. The experimental results show that the proposed method uses Welch spectrum to improve the availability of frequency domain information, which makes the network perform interference recognition more effectively in complex environments.