A BP Neural Network-Based Communication Blind Signal Detection Method With Cyber-Physical-Social Systems

The development of wireless communication technologies will lead to an increase of rapid-time-variants featured occasions and physical occasions of small-data-size communication in network, and the modulation system diversification resulting from social group densification. Though traditional blind signal detection approaches have been improved by reducing their excessive reliance on data size, they still have the defects of various postulated conditions in new settings, complicated computation, unstable detection effect and high error rate. This paper aims to explore a communication blind signal detection method in the setting of cyber-physical-social systems in the field of back-propagation (BP) neural network. First, a BP neural network is used as the equalizer, an error function is redefined, and the back-propagation algorithm is used to train and adjust blind signal data deviation and to update network weight rules for adaptation to network settings with rapid time-variants. Second, a double-sigmoid BP neural network excitation function is constructed to improve poor multiple information processing and network performances resulting from social group densification. Third, self-adaptive variable step size for physical devices' power is constructed to adjust the conflict between convergence rate and steady-state error caused by different powers with the increase of small-data-size communication. Finally, an output vector is made to be closest to an expected vector; thereby, blind signal detection rate of social groups' communication using physical devices in the cyber environment can be improved. The experiments show that the communication blind signal detection method proposed in the paper improves signal detection precision, and reduces omission rate (below 2%). Besides, the method is characterized by minor error, removes deficiencies of traditional methods for blind signal extraction, and can effectively accomplish blind signal detection.