分布式振动光纤信号识别算法研究

Signal recognition algorithm for distributed vibration fiber system

分布式振动光纤周界预警系统理论分析和工程经验表明其灵敏度极高,易受环境干扰,如何降低系统虚警区分振动信号是人为入侵还是环境干扰是实际应用的关键。依据系统实际应用的信号特点,提出了一种基于时域和空间域相结合的预处理算法以及频域分布特征的支持向量机（SVM）算法的两级判别方法。预处理算法利用时域和空间域特征寻找局部空间中的异常振动信号提取可疑事件,SVM算法利用频域特征,识别入侵信号。实验结果验证了所提算法的正确性和有效性。

Theoretical analysis and engineering experience of distributed vibration fiber perimeter pre-warning system shows that the sensitivity is extremely high,and it is vulnerable to environmental interference.In practical applications,it is the key to how to reduce the false alarm rate and distinguish whether the vibration signal is a human or environmental interference.Relying on the signal characteristics of the system in practical application,we propose a two-level identification method based on preprocessing algorithm combining time domain and space domain and support vector machine（SVM）algorithm with frequency domain distribution features.The preprocessing algorithm uses temporal and spatial domain features to find abnormal vibration signals in local space to extract suspicious events and the SVM algorithm uses frequency-domain features to identify intrusion signals.The validity of the proposed algorithm is proved by experimental results.