Same Frequency Interference Detection In Radio Signal Based On Convolutional Neural Network And Support Vector Machine

With the rapid development of radio industry in China,radio interference has become a very serious problem in the process of communication transmission.The traditional algorithm for detecting radio interference relies heavily on prior knowledge and the professionalism of staffs.Aiming at the problem of same frequency interference which is more prominent in radio interference,deep learning is applied to interference signal detection.A new method for detecting same frequency interference in frequency modulation broadcasting is proposed.Firstly,a data set for training deep learning algorithm is produced.We use the signal generator to emit interference signals,and then receive the in-phase and quadrature components of broadcast signals and interference signals by the signal receiver.Then,these signals are restored in MATLAB and the time-frequency images which can reflect the features of the signals are drawn by using time-frequency analysis algorithm.In recent years,the convolutional neural network has made great progress in image recognition.Depending on the characteristics of the data set,we design a seven-layer convolutional neural network as interference detection model.The data set is input into the convolutional neural network and the convolutional neural network is trained to learn the time-frequency features of the signals.The model is optimized by analyzing the training results.Finally,the signal features extracted by convolutional neural network are used to classify the time-frequency images,so as to detect whether the same frequency interference exists in the broadcast signal.The experimental results show that the model can achieve a high accuracy on the validation set in this thesis,which proves the effectiveness of our algorithm and the algorithm can accurately detect same frequency interference in the broadcast signal.Compared with other network models,our model can improve the training speed and shorten the training time under the condition of guaranteeing the accuracy.At the same time,we combine deep learning with traditional machine learning and propose another algorithm to detect same frequency interference.In this thesis,the convolutional neural network is used to extract the time-frequency features of signals,and these features are used as a data set to train the traditional machine learning algorithm.Compared with other traditional machine learning algorithms,support vector machine(SVM)can achieve a better classification result.We also extract the histogram of oriented gradient and gray-level co-occurrence matrix of time-frequency images and combine them as feature data set.Meanwhile,the short-time Fourier transform of the signals is used as another feature set.The advantages and disadvantages of the convolutional neural network in the interference detection task are analyzed by comparing the confusion matrices of support vector machines trained by the three feature sets.The experimental results show that the accuracy of the SVM trained by the features which extracted by convolutional neural network is higher than that of the SVM trained by the other two data sets.Compared with using traditional machine learning algorithm alone,the algorithm combining convolutional neural network with support vector machine in this thesis can accomplish the classification task more conveniently and accurately.