Purification of mixed optical fiber intrusion signal by simplex volume analysis

Abstract

Optical fiber intrusion signal detection is an effective long range technology for perimeter intrusion behavior perception. The optical fiber pre-warning system (OFPS) can detect and recognize many kinds of intrusion signals. However, the current detection-recognition process does not have the purification ability for mixed signals, since the conventional system lacks the unmixing process. The first problem in the unmixing process is how to extract the pure intrusion signal. For this problem, by studying the power spectrum characteristics of intrusion signal, we introduce the statistical volume as the statistical characteristics of intrusion signal in the high dimension space and put forward a novel simplex volume analysis purification method. In the high-dimension space, the mixed signal is located in a simplex composed of the characteristics of the pure signal. By finding the maximum of simplex volume, we can obtain the pure signal of simplex. Furthermore, in order to reduce the interference of outliers as much as possible, we introduce a subspace projection mechanism to remove the outliers. The real data experiments verify the effectiveness of this algorithm as well as good robustness to the noise.