周界入侵监控系统中振动传感器所采集的振动入侵信号中包含环境产生的各种干扰信号，要从这些信号中分离出源振动入侵行为的成分具有重要的现实意义。利用盲信源分离技术来识别源振动入侵信号是一个重要且具有挑战性的学术研究领域。本文在深入研究和比较几种典型的盲信源分离技术的基础上，研究适用于无线周界入侵监控系统及其振动传感器采集的入侵信号的识别方法。主要工作包括： 在研究和比较几种典型的盲信源分离技术的基础上，实现了ICA方法的四种典型算法。实验数据采用自主研发的无线周界入侵监控系统中振动传感器所采集的振动入侵信号。评价指标为相似系数和性能指数。经过实验对比显示：基于JADE算法的ICA方法优于基于Infomax算法、FastICA算法和SOBI算法的ICA方法。因此，基于JADE算法的ICA方法适合用于无线周界入侵监控系统中振动传感器所采集的振动入侵信号的识别。 探索了基于OIE和PSO盲信源分离算法的振动入侵信号识别方法，简称OP-BSS(BSS based on OIE and PSO)。将重叠信息熵(OIE)作为盲信源分离算法中的目标函数，粒子群优化算法(PSO)作为对目标函数的优化算法。实验数据分别采用人工合成数据和自主研发的无线周界入侵监控系统中振动传感器所采集的振动入侵信号。实验结果表明，该方法可以有效地识别出源振动入侵信号。 基于上述研究结果，将振动入侵信号识别方法应用于自主研发的无线周界入侵监控系统中。系统包含网络节点、路由节点、中心节点和上位机报警监控软件四大模块。测试结果表明该系统功能正确，达到了预期的目标。

Vibration sensor’s signals are collected from perimeter intrusion monitoring system, which include a variety of interference signals caused by environment. It is important and practical to separate the source of vibration intrusion signals from those signals. Using blind source separation (BSS) technique to recognize the source of vibration intrusion signals is an important and challenging research field. With careful study and comparison with several typical blind source separation methods, recognition methods are introduced to adapt to wireless perimeter intrusion monitoring system and its vibration sensor’ signals. The main contributions are as follows: With careful study and comparison with several typical blind source separation methods, four kinds of typical algorithms based on ICA are implemented in this thesis. Experimental data are vibration sensor’ intrusion signals, which are collected from wireless perimeter intrusion monitoring system. Similarity coefficient and performance index are used to evaluate the performance of four kinds of algorithms. Comparing JADE algorithm with the Infomax algorithm, FastICA algorithm and SOBI algorithm respectively, which are all based on ICA algorithm. Experiments show that JADE algorithm is better than other three algorithms. Therefore, JADE algorithm is suitable for recognition of vibration sensors’ intrusion signals, which are collected from wireless perimeter intrusion monitoring system. A blind source separation algorithm named OP-BSS is proposed, which is based on Overlap Information Entropy (OIE) and Particle Swarm Optimization (PSO). The OIE is an objective function in blind source separation algorithm, and PSO is used as optimization algorithm. Experimental data are synthetic data and vibration sensor’ intrusion signals severally. The experimental results show that this method can effectively recognize the source of vibration sensors’ intrusion signals. On the basis of above research, identification methods of vibration intrusion signals are applied to wireless perimeter intrusion monitoring system, which is developed independently. The system contains four modules, which are network nodes, routing nodes, central node and upper computer alarm monitoring software. The test results show that the system functions are correct, and the desired goal is achieved.