Using Deep Convolutional Neural Network to Recognize LTE Uplink Interference
Jie Ren, Xin Zhang, Yan Xin · 2019
In this paper, we provide a novel and field-proven approach to recognize Long Term Evolution (LTE) uplink interference. We first pre-process the time domain signal into spectral waterfall so that we can formulate the LTE uplink interference recognition problem as an image classification task. We design a convolutional neural network (CNN) based learning system to solve the problem. We then show by experimental studies that the proposed CNN model achieves a 95% accuracy on average even with imbalanced and limited training data. Furthermore, we evaluate the proposed CNN model by visualizing the learned parameters of the model. We find that local patterns of the uplink interference can be well learned through the first layer convolution filters of the model. To the best knowledge of the authors, this paper represents the first application of CNN to LTE uplink interference recognitions.