An efficient automatic modulation recognition using time–frequency information based on hybrid deep learning and bagging approach

Abstract

Determining the type of modulation is an important task in military communications, satellite communications systems, and submarine communications. In this study, a new digital modulation classification model is presented for detecting various types of modulated signals. The continuous wavelet transform is used in the first step to create a visual representation of the spectral density of the frequencies of the modulation signals in a scalogram image. The subsequent stage involves the utilization of a deep convolutional neural network for feature extraction from the scalogram images. In the next step, the best features are chosen using the MRMR algorithm. MRMR algorithm increases the classification speed and the ability of interpret the classification model by reducing the dimensions of the features. In the fourth step, the modulations are classified using the group learning technique. In the simulations, modulated signals with different amounts of noise with SNR from 0 to 25 dB are considered. Then, accuracy, precision, recall, and F1-score are used to evaluate the performance of the proposed method. The results of the simulations prove that the proposed model with achieving above 99.9% accuracy performs well in the presence of different amounts of noise and provides better performance than other previous studies.