TY - GEN
T1 - A Deep Learning Method Based on Convolution Neural Network for Blind Demodulation of Mixed Signals with Different Modulation Types
AU - Zhu, Hongtao
AU - Wang, Zhenyong
AU - Li, Dezhi
AU - Guo, Qing
AU - Wang, Zhenbang
N1 - Publisher Copyright:
© 2019, ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering.
PY - 2019
Y1 - 2019
N2 - In recent years, deep learning is becoming more and more popular. It has been widely used in image recognition, automatic speech recognition and natural language processing. In the field of communication, the signal is considered as time data, which can identify the intrinsic characteristics and information of the signal by the way of deep learning. In the aspect of cognitive radio, if the signal adopts different modulation methods in different time slots for adaptive modulation, it is difficult for the existing signal demodulation system to demodulate it effectively. Usually, it is necessary to identify the modulation mode of the signal first. In this context, the deep learning is introduced into signal demodulation. On the basis of analyzing the structure of convolutional neural network (CNN), an improved CNN structure is proposed, which does not need to recognize modulation methods and realizes blind demodulation of mixed signals with different signal-to-noise ratio (SNR). Through transfer learning and denoising auto-encoder, the network is further optimized to further reduce the bit error rate (BER).
AB - In recent years, deep learning is becoming more and more popular. It has been widely used in image recognition, automatic speech recognition and natural language processing. In the field of communication, the signal is considered as time data, which can identify the intrinsic characteristics and information of the signal by the way of deep learning. In the aspect of cognitive radio, if the signal adopts different modulation methods in different time slots for adaptive modulation, it is difficult for the existing signal demodulation system to demodulate it effectively. Usually, it is necessary to identify the modulation mode of the signal first. In this context, the deep learning is introduced into signal demodulation. On the basis of analyzing the structure of convolutional neural network (CNN), an improved CNN structure is proposed, which does not need to recognize modulation methods and realizes blind demodulation of mixed signals with different signal-to-noise ratio (SNR). Through transfer learning and denoising auto-encoder, the network is further optimized to further reduce the bit error rate (BER).
KW - Convolution neural network
KW - Denoising auto-encoder
KW - Signal demodulation
KW - Transfer learning
UR - https://www.scopus.com/pages/publications/85065907344
U2 - 10.1007/978-3-030-19153-5_9
DO - 10.1007/978-3-030-19153-5_9
M3 - 会议稿件
AN - SCOPUS:85065907344
SN - 9783030191528
T3 - Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST
SP - 91
EP - 103
BT - Wireless and Satellite Systems - 10th EAI International Conference, WiSATS 2019, Proceedings
A2 - Jia, Min
A2 - Guo, Qing
A2 - Meng, Weixiao
PB - Springer Verlag
T2 - 10th EAI International Conference on Wireless and Satellite Systems, WiSATS 2019
Y2 - 12 January 2019 through 13 January 2019
ER -