基于labview的信号拟合算法研究-控制理论与控制工程专业论文.docxVIP

基于LabVlEw的信号拟合算法研究Abstract 基于LabVlEw的信号拟合算法研究 Abstract For s如挈1血fitting，lots of researches on neural networks have been done in也e applications a11d the theories．The劬ctions palette in LabVIEW provides many algorithms of linear aIld non．1inear of sigIlal 6ttillg．We caIl use releVaIlt methods to signal processing accordjng to di＆rent signaJs and they have adVaJltages柚d disadvantages of their own．To oVercome t|le drawbacks of t11e current metllods，people need to咖dy continually Tb improVe the accuracy of signal fitting by embeddiIlg tlle improved algorithm into LabVIEW，we do some study and jmproV咖ent on algorithm ofsignal fit亡ing by usjng SoR Function Basis aIld Simulated A叽ealing Algorithm in“s paper． FirStly，study on soft如nction baSis ofneural networks in t11e papeL Although，Neural netWork has mally applications，it still haS some problems，one of which is neuron function expression of n“ible and chaIlgeful is few a11d moSt function fbm is simple，such as Sigmoid mnction、radial mnction．In order to solVe tKs problem，we employ other nnctions，such as soR fbction baSis wi山M par啪eters．By adjusnnent of parameters， signals fitting can reach originaJ signaIs． Generall弘the moSt coⅡmlon used met}lod for the adjustment is the Fadient descent． The method has many advalltages，for ex锄ple，its speed of training is fbt aIld its appIications are conVenient and extensive．Howeve r，it has drawback：jt’s easy to fall into the 10cal minim岫during t11e trainjng Actually most of the problems have many extrema，and the gradient descending must be repeatedly used in the signal fitting process．When廿1e searches are over，、^，e、vould select the best extremum．In order to accelerate the search，we ma：ke use of a global algorithm：simuJated锄ealingalgorit№． Based on simuIated arulealing aIgorithm(SAA)aIld so代function basis(SFB)， algorithm of deallng with one dimensional signal fitting by means of neural network is presented．The combination of SFB aIld SAA and t}le selection of soR SFB are analyzed in emphasis．S