基于深度学习的QPSK智能接收机模型研究

Design of QPSK Intelligent Receiver Based on Deep Learning

摘要:

针对通信信道中存在噪声等干扰因素时，QPSK接收机解调接收信号性能较差的问题，本文研究了一种基于深度学习的QPSK智能接收机模型。该QPSK智能接收机模型由LSTM神经网络和全连接层构成，借助了递归神经网络中的内存结构，也利用了LSTM能提取接收信号的时间相关性这一特点。在信噪比为0~7dB的条件下进行仿真实验，实验结果表明，在加性高斯白噪声，同相和正交失衡以及频率偏差干扰因素影响下，本文所研究的QPSK智能接收机模型在0~7dB时的误码率相比于使用传统硬判决方法的通信接收机的误码率得到了显著降低。其中，QPSK智能接收机模型在7dB时的误码率低至0.0109%，大约只有传统硬判决方法误码率的1/7。在发生频偏和IQ失衡时，QPSK智能接收机模型在7dB时的误码率分别低至0.0147%和0.0198%，都远低于相同条件下传统硬判决方法的误码率。因此，采用本文所研究出的QPSK智能接收机模型能够显著提高接收机的检测性能。

Abstract:

When there are interference factors such as noise in the communication channel, the performance of quadrature phase shift keying receiver in demodulating received signals is poor.In view of the problem,this paper studied a QPSK intelligent receiver model based on deep learning.The QPSK intelligent receiver model is composed of long and short-term memory neural network and fully connected layer. With the help of the memory structure in the recurrent neural network, it also uses the characteristic of LSTM can extract the temporal correlation of the received signal. Simulation experiments with a signal to noise ratio at 0 to 7 dB showed that,under the influence of Gaussian additive white noise,inphase and quadrature imbalance and frequency deviation interference factors, the bit error rate of the proposed QPSK intelligent receiver model at 0 to 7 dB is significantly reduced compared with that of the communication receiver using the traditional hard decision method.Among them, the bit error rate of QPSK intelligent receiver model at 7dB is as low as 0.0109%, which is only about 1/7 of the bit error rate of the traditional hard decision method. In the event of frequency deviation and IQ imbalance, the bit error rate of QPSK intelligent receiver model at 7dB is as low as 0.0147% and 0.0198%, respectively, which are much lower than the bit error rate of the traditional hard decision method under the same condition. Therefore, adopting the QPSK intelligent receiver model proposed in this paper can significantly improve the detection performance of the receiver.