Intrusion signal recognition in OFPS under multi-level wavelet decomposition based on RVFL neural network

The performance of the optical fiber pre-warning system (OFPS) is susceptible to the interference of background noise, especially in the low signal-to-noise ratio (SNR) environment. In this paper, a recognition method based on the multi-level wavelet decomposition is proposed to accurately identify the running and digging intrusion signals in OFPS under the low SNR condition. The method includes the cross-correlation operation, feature extraction, and recognition using a random vector functional-link (RVFL) neural network (NN). Firstly, the collected signals are processed by the cross-correlation operation. Compared with the conventional filtering method, the background noise can be suppressed to the greatest extent by the cross-correlation operation, which keeps the signal details as much as possible. Secondly, the cross-correlation functions obtained from the cross-correlation operation are decomposed into five levels by the multi-level wavelet decomposition. And then the average energy ratios can be obtained along with the decomposed levels, and we select these ratios in the five frequency bands as the feature of intrusion signals. Next, the feature samples are sent into the RVFL NN for training so as to complete the recognition of these intrusion signals. Finally, the effectiveness of the algorithm is verified by the field experiment.