Electronic Signal Processing and Pattern Recognition Technology Based on Deep Learning

Abstract

Aiming at the challenge that traditional methods can not meet the requirements of high precision and high efficiency in complex scenes, a method of introducing deep learning model to improve the accuracy and efficiency of signal processing is proposed. Specifically, this paper designs an improved convolutional neural network model named "signal recognition convolutional network" (SRCN), which is optimized on the basis of traditional CNN to better adapt to the characteristics of electronic signals. In SRCN model, the local features of the signal are automatically extracted in the convolution layer through one-dimensional convolution kernel, and the dimension of the feature map is reduced by maximum pooling, and finally the classification results are output through softmax function in the fully connected layer. The SRCN model is trained and verified by using 30,000 data sets including radar signals, communication signals and other types of electronic signals. The results show that the SRCN model has a high classification performance on all kinds of signals, with an accuracy rate of 95.0%, which is significantly better than the traditional support vector machine (SVM) and decision tree methods. The research in this paper provides new ideas and methods for the field of electronic signal processing and pattern recognition, and shows the great potential and application value of deep learning technology in this field.