1 / 11

Tapering and prewhitening fFT

Tapering and prewhitening fFT. t aper, h(u). Need for prewhitening / prefiltering periodogram is generally biased. Smoothed periodogram. Expected value generally biased, but not for white noise. Improve by prefiltering Y filtered version of X with transfer function A

emilie
Télécharger la présentation

Tapering and prewhitening fFT

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Tapering and prewhitening fFT taper, h(u)

  2. Need for prewhitening/prefiltering periodogram is generally biased

  3. Smoothed periodogram Expected value generally biased, but not for white noise

  4. Improve by prefiltering Y filtered version of X with transfer function A dZY(λ) = A(λ)dZX(λ) fYY(λ) = |A (λ)|-2 fXX(λ) fXX(λ) = |A (λ)|-2fYY(λ) For example: detrend, “remove” peaks, fit autoregressive

  5. The importance of prefiltering. The estimates are generally biased E{ fT()} =  WT (-) f() d Seismic noise horizontal components recorded at UCB Estimated coherence can be 1,0 when no relation Estimated coherence can be 0.0 when linearly time invariantly related

  6. postscript(file="seismic.eps") junkx<-scan("ts_drb.16.dat") junky<-scan("ts_drb.17.dat") par(mfrow=c(2,1)) xaxis<-c(1:2500)*.08 plot(xaxis,junkx,type="l",main="Vertical component seismic noise at Berkeley",xlab="time (sec)",ylab="",las=1) plot(xaxis,junky,type="l",main="West component",xlab="time(sec)",ylab="",las=1) par(mfrow=c(2,3)) junk<-spec.pgram(cbind(junkx,junky),spans=15,taper=0,detrend=F,demean=T,plot=F) junk$freq<-junk$freq/.08 plot(junk$freq,junk$spec[,1],type="l",las=1,main="Vertical noise",xlab="frequency (hertz)",log="y") plot(junk$freq,junk$spec[,2],type="l",las=1,main="West noise",log="y") plot(junk$freq,junk$coh,type="l",ylim=c(0,1),main="Raw data",las=1) abline(h=1-(1-.95)^(1/(.5*(junk$df-2))))

  7. junkxx<-ar(junkx,order.max=2) junkyy<-ar(junky,order.max=2) Junkx<-junkxx$resid Junky<-junkyy$resid Junkx<-Junkx[3:length(Junkx)];Junky<-Junky[3:length(Junky)] Junk<-spec.pgram(cbind(Junkx,Junky),spans=15,taper=0,detrend=F,demean=T,plot=F) Junk$freq<-Junk$freq/.08 plot(Junk$freq,Junk$spec[,1],type="l",las=1,log="y") plot(Junk$freq,Junk$spec[,2],type="l",las=1,log="y") plot(Junk$freq,Junk$coh,type="l",ylim=c(0,1),main="AR(2) residuals",las=1) abline(h=1-(1-.95)^(1/(.5*(junk$df-2)))) graphics.off()

More Related