Recognition Of Construction Equipment Based On OSALPCC Features And SVM

With the rapid development of economy,the electricity-consumption is increasing year by year.In the power distribution system,the proportion of underground cables is becoming higher and higher,because the underground cables has advantages of little occupied area,high reliability of power supply and will not affected by environment.But in the process of power supply,the underground cables often subjected to external damages(mainly come from 4 kinds of construction equipment: the impact hammer,cutting machine,grab excavator and hydraulic percussion hammer),which will influence the reliability of power supply.Thus,it is a crucial technical problems for electric power sector to protect the underground cables from external damages.Nowadays,the electric power sector protect the underground cables from external damages mainly by artificial patrols,set warning signs or distributed optical fiber sensor system.But,all these measures needs a lot of manpower and material resource.The distributed optical fiber sensor system have the disadvantage of high false alarm rate,and cannot identify the vibration source.Because of all the reason above,this paper brings out a new method based on acoustics to protect the underground cables from external damage.This paper introduced the process of the construction equipment recognition,including the acoustic feature extraction algorithms of construction equipment and the intelligent classification algorithms based on support vector machine(SVM),then compared the effect with k-nearest neighbor(KNN).Finally,we developed a software to monitor the underground cables by detect the acoustic signal generated by construction equipment.What this paper investigate is shown as bellow:1)The collect and pretreatment of acoustic signals.Including the collect of the acoustic signal,the pre-emphasis,framing of signal.2)Analysis and feature extraction of the acoustic signal.Analyzed the acoustic signal of construct equipment,modify the feature extraction algorithm(Linear Prediction Cepstrum Coefficient,LPCC)of speech to classify the construction equipment.But the algorithm shows low performance in case of noisy environment,so,we proposed the developed algorithm OSALPCC(One-sided Autocorrelation Linear Prediction Cepstrum Coefficient),we compared the difference of the feature vector for different equipment acoustic signals,and come to the conclusion that the construction equipment can be classified by the feature we proposed.3)Intelligent classification algorithm.This section introduced the basic concepts of support vector machine(SVM),and compared the recognize rate with the feature vector of LPCC and OSALPCC.Through the experiment we found that both feature vector shows good performance.But the OSALPCC feature performs better than LPCC under the circumstance of noisy environment.4)Software development.Developed a monitor software to protect underground cables from external damage based on the platform of LabVIEW,according to the experiments,the system performs better,the test accuracy can reach more than 90%.