基于SVD和MPSO-SVM的光纤周界

针对光纤振动信号有噪声干扰、识别信号类型准确率不高且识别时间长的问题,提出了基于奇异值分解(singular value decomposition, SVD)和改进粒子群优化支持向量机(modified particle swarm optimization support vector machine, MPSO-SVM)的识别方法。首先,采用SVD对信号去噪,根据奇异值序列二阶差分谱单边极小值原则确定信号重构秩阶次。其次,提取振动信号特征,利用串行特征融合(serial feature fusion, SFF)方法组建特征向量组。最后,利用MPSO-SVM进行分类识别,提高识别精度和算法效率。采用实测信号进行验证,结果表明,信噪比有明显提升,信号平均识别率较粒子群优化支持向量机(particle swarm optimization support vector machine, PSO-SVM)提升5%。该方法较传统神经网络识别方法有较好的效果,具有实际应用价值。

Optical fiber perimeter vibration signal recognition based on SVD and MPSO-SVM

A recognition method based on singular value decomposition (SVD) and modified particle swarm optimization support vector machine (MPSO-SVM) is proposed for the optical fiber vibration signals with noise interference, low accuracy and long recognition time. Firstly, SVD is used to denoise the signal, and the rank order of signal reconstruction is determined according to the single-side minimum principle of second-order difference spectrum of singular value sequence. Secondly, the vibration signal features are extracted and a set of feature vectors is constructed by means of serial feature fusion (SFF). Finally, MPSO-SVM is used for classification and recognition to improve the accuracy and efficiency of the algorithm. The measured signal is used for verification. The results show that the signal to noise ratio is significantly improved, and the average recognition rate is 5% higher than that of PSO-SVM. This method performs better than the traditional neural network recognition method and has the practical application value.