Asynchronous Wireless Signal Modulation Recognition Based on In-Phase Quadrature Histogram

Abstract

Automatic modulation recognition is a key technology in the field of signal processing. Conventional recognition methods suffer from low recognition accuracy at low signal-to-noise ratios (SNR), and when the signal frequency is unstable or there is asynchronous sampling, the performance of conventional recognition methods will deteriorate or even fail. To address these challenges, deep learning-based modulation mode recognition technique is investigated in this paper for low-speed asynchronous sampled signals under channel conditions with varying SNR and delay. Firstly, the low-speed asynchronous sampled signals are modeled, and their in-phase quadrature components are used to generate a two-dimensional asynchronous in-phase quadrature histogram. Then, the feature parameters of this 2D image are extracted by radial basis function neural network (RBFNN) to complete the recognition of the modulation mode of the input signal. Finally, the accuracy of the method for seven modulation methods is verified by extensive simulations. The experimental results show that under the channel model of additive white Gaussian noise (AWGN), when the SNR of the input signal with low-speed asynchronous sampling is 6 dB, more than 95% of the average recognition accuracy can be achieved, and the effectiveness and robustness of the proposed scheme are verified by comparative experiments.