Research On Vibration Signal Recognition Method For Pedestrian Movement

Theft is a common crime of multiple financial invasion.However,due to people’s weak awareness of prevention,insufficient anti-theft measures and sophisticated criminal means,the property security of the state,enterprises and the broad masses of the people is seriously threatened,which seriously affects people’s normal life.Therefore,people are in urgent need of a security detection technology with high accuracy and real time,so as to provide important information for security prevention work.Based on this,vibration signal recognition theory and technology has become an important research direction in the field of security.In this paper,vibration signal recognition technology is studied based on vibration signals generated by different pedestrian movement modes.Firstly,the paper introduces the generation mechanism of ground vibration signal,and uses a seismic detector and oscilloscope to build a vibration signal acquisition device,which collects the ground vibration signals caused by 9 different states,including outdoor natural noise,outdoor single walking,outdoor double walking,outdoor triple walking,outdoor single running,indoor single walking,indoor double walking,indoor triple walking and indoor single running,The sample database is built.Then,this paper adopts the wavelet threshold denoising algorithm to de-noise the collected sample signals,then extracts the time-domain and frequency-domain features of the signals,and takes the energy of each intrinsic modulus function(IMF)component of the signals obtained after empirical wavelet transform as the features,thus constructing the feature vector,which lays the foundation for the subsequent classification and recognition.Finally,this paper adopts support vector machine(SVM)algorithm and onedimensional convolutional neural network(1DCNN)algorithm respectively to classify and recognize vibration signals of pedestrian movement.In this paper,a pedestrian vibration signal recognition method based on particle swarm optimization(PSO)algorithm is proposed to optimize SVM,and the average recognition accuracy is94.00%.However,SVM algorithm requires researchers to carry out complex and tedious feature extraction,and the classification accuracy of the model depends on the quality of the extracted features.Therefore,this paper proposes a vibration signal recognition method for pedestrian movement based on 1DCNN.In this paper,we first proposed a vibration signal recognition method based on the classical 1DCNN model for pedestrian movement.After experimental verification,the average recognition accuracy of 94.96% is obtained,and the average prediction time of the model is 3.27 ms.Then,in order to make the model more real-time and practical,a vibration signal recognition method based on the deep convolutional neural networks with wide firstlayer kernel(WDCNN)model is proposed.The experimental results show that the average recognition accuracy rate of 94.52% is obtained by using the WDCNN model to classify and recognize vibration signals of pedestrian movement,which is not much different from that obtained by using the classical 1DCNN.However,the number of parameters of the WDCNN model is greatly reduced,the training time of the model is greatly reduced,and the average single prediction time of the model is only 0.254 ms,which meets the requirements of high accuracy and strong real-time performance.