RIDGETRIM Trim edge effect regions from wavelet ridges. Wavelet ridges, such as those computed by RIDGEWALK, are generally contaminated by edge effects at both ends. In these locations, the ridges give biased estimates of the signal properties. FRO=RIDGETRIM(P,RHO,FR), where P=SQRT(BETA*GAMMA) and FR is a column vector of frequencies from the ridge analysis, removes RHO*P/pi periods from the beginning and end of each ridge. Ridges containing less than two points after this trimming are removed. Note that P=SQRT(BETA*GAMMA) is characterizes the generalized Morse wavelet used in the wavelet transform. P/pi is approximately the number of oscillations within the central window of the wavelet, see eqn. (37) of Lilly and Olhede (2009). The choice RHO=1 thus corresponds to trimming one wavelet half-width from end. Numerical experiments show RHO=1 is a good value, but this depends somewhat on the values of GAMMA and BETA chosen. For joint ridges for multivariate signals FR should be a column vector of the joint instantaneous frequency. This is formed from the array FR output by RIDGEWALK by the weighted average FR=VMEAN(FR,1,ABS(WR).^2). RIDGETRIM(DT,P,RHO,FR) specifies that FR has been computed with the sample rate DT. [FRO,X1O,X2O,...,XNO]=RIDGETRIM(P,RHO,FR,X1,X2,...,XN), where the XN are any other ridge quantities, trims these in the same way as FR. The XN are arrays having the same number of rows as FR. See also RIDGEWALK. 'ridgetrim --f' generates a sample figure. Usage: fr=ridgetrim(P,rho,fr); [fr,wr,ir,jr]=ridgetrim(P,rho,fr,wr,ir,jr); [fr,wr,ir,jr]=ridgetrim(dt,P,rho,fr,wr,ir,jr); __________________________________________________________________ This is part of JLAB --- type 'help jlab' for more information (C) 2018 J.M. Lilly --- type 'help jlab_license' for details