One of the most challenging problems in unattended ground sensor (UGS) systems is to distinguish human footstep signals from noise sources. As the distance between the sensor and the moving target increases, the signal-to-noise ratio (SNR) decreases rapidly. Many methods have been proposed to solve this problem, but most of them suffer from unacceptably high false alarm rate or omissive report rate in practical applications. In this paper, a novel approach based on parallel recurrent neural network (PRNN) is proposed to improve the seismic target recognition performance. The PRNN is composed of a time domain feature network and a frequency spectrum feature network. The time domain feature network is used to handle the running signals, and the frequency spectrum feature network is used to handle the walking signals.The output of the PRNN is a fusion of the two networks. Experimental results show that the proposed approach can improve the human recognition accuracy up to 98.3% and has a remarkable performance compared with other machine learning methods.INDEX TERMS Data fusion, parallel recurrent neural network, signal-to-noise ratio, unattended ground sensor.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.