关键词: 多输入多输出信号, 信噪比估计, 调制识别, 神经网络, 多任务学习

Abstract: Aiming at the problem that the current research on Blind SNR estimation and subcarrier modulation recognition of MIMO-OFDM signal in non cooperative communication only focuses on a single task, an algorithm combining deep neural network and multi task learning (MTL) framework is proposed to complete blind SNR estimation and modulation recognition at the same time. Firstly, the joint approximate diagonalization algorithm (JADE) of eigenvalue matrix is used to recover the transmitted signal, and the codirectional orthogonal (I/Q) component of the recovered signal is extracted as the shallow feature; Then, a multi task learning model based on one-dimensional convolutional neural network (CNN) is built to realize complementary advantages through joint training of signal-to-noise ratio (SNR) estimation and modulation recognition. Simulation results show that the proposed algorithm can achieve better performance than single task learning (STL). When the signal-to-noise ratio is - 10dB, the mean square error of signal-to-noise ratio estimation is reduced by 66.21% and the modulation recognition accuracy is improved by 4.75%. In addition, when the signal-to-noise ratio is greater than - 1dB, the mean square error of signal-to-noise ratio estimation is less than 0.1; When the signal-to-noise ratio is 3dB, the accuracy of modulation recognition can reach 100%.

Key words: Multiple Input Multiple Output (MIMO) signals, SNR estimation, Modulation identification, Neural network, Multi task learning