Recognition and prediction of ground vibration signal based on machine learning algorithm
- 57 Downloads
Accurate recognition of the type of ground motion is a basic task in the field of seismic engineering. In this paper, the key technologies of detection and recognition of underground seismic signal are studied. The target recognition algorithm is designed to realize the target recognition through denoising the collected target signal and extracting the characteristic quantity. Considering that ground motion signals generated by moving targets on the ground are susceptible to environmental noise, this paper introduces the working principle of wavelet packet denoising and its support vector machine classification model. Wavelet packet was used to transform the signal to denoise first, then zero-crossing rate analysis of the denoised signal was carried out after wavelet packet denoising and extracts the parameters, and the energy of cross-correlation criteria was selected finally. Quantitative indices are combined to construct multi-feature vectors, which are used as input of multi-class support vector machine for training and prediction. In this model, the optimal parameters of support vector machine model are found by genetic algorithm parameter optimization. The experimental results show that the improved model can recognize and classify the ground motion signals caused by people and vehicles correctly and can improve the performance of the classifier.
KeywordsGround motion signal Wavelet packet transform Support vector machine Genetic algorithm Target recognition
This work was supported by Jilin Province Science and Technology Development Plan Project (Grant No. 20180201036GX).
- 5.Zhou F, Gan J, Lv H, Cui L (2018) Application of wavelet analysis in underground embedded distributed optical fiber vibration monitoring system. In IOP conference series: earth and environmental science, vol 189, no 3, p 032025, IOP PublishingGoogle Scholar