The Research On Vibration Source Identification Algorithm Based On Multidimensional Characteristics

Vibration signal recognition is the important part of optical fiber pipeline safety early warning system. In order to solve the safety early warning system of vibration signal classification problem, when the source of vibration signal processing, it need to extract the source signal characteristic. Source of vibration signal recognition rate of high and low, depending on the source of vibration signal feature extraction of robustness and accuracy. Therefore, the source of vibration signal feature extraction in the source of vibration signal classification plays an important role.First of all, the paper studies the method of the source of vibration signal of detected data preprocessing,it introduces the frame was carried out on the vibration signal processing methods, at the same time,the paper studied the principles and methods of wavelet transform, wavelet denoising method.Secondly, through different event behavior characteristics of the source of vibration signal analysis and research the dangers to the main characteristics of the pipeline vibration events, including time domain analysis, frequency domain analysis and time frequency analysis,mainly studied four properties characteristics of vibration signals, these four kinds of signals are the picks plane, car, electric drills and electric pick. Mainly from the duty ratio, pitch frequency, energy eigenvector, center frequency and so on to extract the vibration signal characteristic parameters, establish a multidimensional eigenvector and vibration signal characteristic library.Finally,the paper studies the support vector machine in the application of optical fiber source identification strategy, mainly study the basic model,deeply understanding of the feature space interval on the biggest significance of linear classifier, deeply research for solving the quadratic programming problem, at the same time, the selection of the structure of the kernel function and parameters for the study of theory, analyses the SMO iteration algorithm, the use of the minimum sequence optimization algorithm for training the SVM classifier, multiple binary model was set up, using the extracted vibration source data of multidimensional feature vector through the classifier to classify recognition. The test results show that the accuracy is higher, by using the method to identify fiber optic vibration signal, it is provide a solution for identification of the optical fiber vibration signal.