论文部分内容阅读
针对M-Z干涉仪型光纤分布式扰动传感系统输出信号短时频率随外界扰动变化的特征,提出了基于短时频率-时间特性的模式识别算法。采用提取短时过电平率来描述传感信号的短时平均频率-时间特性,并将提取出来的时频特性分段后建立相应的特征元素模型,通过动态规划算法(DTW)筛选出最优特征元素模型,将信号所有最优模型的参数作为信号特征输入到人工神经网络(ANN)进行学习和判决,降低了ANN的训练难度以及对时间的敏感性,提高了系统的环境适应能力。实验结果表明:该方法可以有效区分瞬时作用、长时作用、径向作用和不规则作用等多种不同扰动事件,平均识别速度在0.26 s之内,平均识别准确度在97%以上。
Aiming at the feature that the short-time frequency of the output signal of M-Z interferometer optical fiber distributed disturbance sensor system varies with disturbance, a pattern recognition algorithm based on short-time frequency-time characteristics is proposed. The short-time average over-time rate is used to describe the short-term average frequency-time characteristics of the sensing signal. The extracted time-frequency characteristics are segmented to establish the corresponding eigen element model, and the dynamic programming algorithm (DTW) The optimal feature element model inputs the parameters of all the optimal signal models into the artificial neural network (ANN) for signal learning and judgment, which reduces the training difficulty and time sensitivity of the ANN, and improves the environment adaptability of the system. The experimental results show that the proposed method can effectively distinguish different disturbance events such as instantaneous action, long-term action, radial action and irregularity. The average recognition speed is within 0.26 s and the average recognition accuracy is above 97%.