Research On The Identification Of Personnel Based On The Time Difference Between The Left And Right Footsteps

With the popularization of intelligence in people's social life,identity recognition technology based on traditional methods such as ID cards,passwords,passwords,and passwords can no longer meet people's needs because of its shortcomings such as easy to lose,easy to forge,and inconvenient.Identity recognition technology based on biometrics such as face,fingerprint,and iris has become a hot spot in the current identity field because of its high security,strong stability,no loss and forgetting.Although the development of these technologies has been relatively mature,there are also certain defects,such as the need for people and acquisition equipment to actively cooperate and get close to each other.Compared with the above identification technology,the identification technology based on the footstep vibration signal has the advantages of better concealment,inescapability and non-contact.When people walk normally,because of their height,weight,personality and other different ways of walking will be different,and then the resulting footstep vibration signal is also different,through the vibration sensor installed on the ground to collect people's normal walking footstep vibration signal,the collected footstep vibration signal analysis,extract can distinguish the characteristics of pedestrian identity and then achieve pedestrian identity recognition.This technology can be applied to smart buildings,public security and health monitoring and other fields.In this paper,a pedestrian number identification method based on the characteristics of the time-frequency domain and a pedestrian identification method based on the difference between the left and right foot vibration signals are proposed,and the specific work is as follows:(1)Combined with the relevant knowledge of seismic engineering,the mechanism of footstep vibration signal generation and propagation is analyzed,and a set of equipment that can clearly collect footstep vibration signals is combined,including 941 B sensor,COINV collector,and laptop computer.According to the Nyquist sampling theorem combined with the characteristics of the footstep vibration signal,it is determined that the sampling frequency of the footstep vibration signal is most appropriate at 1024 Hz.The presence of noise will adversely affect the analysis and identification of the footstep vibration signal,so before the characteristic analysis of the footstep vibration signal,it is necessary to filter the collected footstep vibration signal and reduce noise to improve the signal-to-noise ratio of the foot vibration signal.In this paper,the noise reduction effects of three noise reduction methods of wavelet threshold noise reduction,EMD noise reduction,and EMD-wavelet threshold combined noise reduction are compared,and the signal-to-noise ratio of the signal-to-noise after using the EMD-wavelet threshold combined noise reduction method is significantly higher than that of the other two methods,and more information about the original signal is retained.The presence of noise will adversely affect the analysis and identification of the footstep vibration signal,so before the characteristic analysis of the footstep vibration signal,it is necessary to filter the collected footstep vibration signal and reduce noise to improve the signal-to-noise ratio of the foot vibration signal.In this paper,the noise reduction effect of three noise reduction methods of wavelet threshold noise reduction,EMD noise reduction,and EMD-wavelet threshold combined noise reduction is compared,and the signal-to-noise ratio of the EMD-wavelet threshold combined noise reduction method is significantly higher than that of the other two methods,and more information of the original signal is retained,so the EMD-wavelet threshold joint noise reduction method is used in this paper.(2)Through the analysis of the vibration signal of multiple people's footsteps,a pedestrian number identification method based on the characteristics of the time frequency domain is proposed.Walking with multiple people together will produce a "lock step" mentality,that is,when walking,everyone's stride length and pace frequency will tend to be consistent,so the footstep vibration signal pattern is similar to that of multiple people walking together and when walking alone.In order to identify the number of travelers from this similar pattern,instead of extracting a single footstep vibration signal for analysis,a five-second "useful" data segment is extracted from the data set of pedestrians passing through the sensor,and the five-second data segment is analyzed from the time domain and frequency domain angles,a total of 22 features are extracted,and the 22 features are sorted by random forest algorithm,and the optimal subset is selected in combination with SVM.The data of 300 footstep vibration signal samples of 1-3pedestrians were classified and identified,and the recognition rate of the above proposed method was 100%.(3)Through the analysis of the single-person footstep vibration signal,a method is proposed to make a difference by continuously making a difference between the time curve of the left and right footstep vibration signal to eliminate the influence of shoe type and floor type: centered on the peak point of the time curve of the foot vibration signal speed,Take the time curve(peak curve)of 0.03 s before the peak point and the 0.03 s after the peak point,the time curve(peak curve)with a total duration of 0.06 s,the time curve of0.06 s obtained by the adjacent left and right feet is taken to obtain the time range curve of the left and right footstep vibration,and the average of the 7 consecutive footstep vibration time interval curves(referred to as the average curve)is taken,and the 9 characteristic parameters of the average curve are used as the basis for personnel identification.In order to prove the feasibility of the above method,10 pedestrian identification experiments were carried out in four vibrational environments,and each experiment was to identify 100 sets of sample data of 10 pedestrians,with the minimum recognition rate of 90% and the maximum 96%,thus demonstrating the feasibility of the above method.