f



Interpreting Wavelet Coefficients: Large Coefficients at higher levels

Hello, 

I have a high resolution time series of measurements, taken throughout a 24 hour period, that I would like to analyze on different time scales i.e look at the magnitude of the 1 sec variations, 5 sec variations, minute, hour etc.  I figured I would use a the discrete wavelet transform and look at the different levels which should be on time scales of (2^N )*dt, where dt is my sampling rate of 10 ms. 

My data is normalized to 1 and I am seeing very large (>100) magnitude coefficients at higher wavelet levels and I was unsure if this was expected or I am doing something incorrectly.  

For example, 
% s is normalized time series data
[c,l] = wavedec(s,20,'db4');

l(1), l(2) for ca20,cd20 = 4,4 

and c(1:8) = 
  641.3052
  623.5509
  260.5610
  165.5240
  -41.9139
   44.4042
   99.4153
         0

I will admit I am fairly new to wavelets, so this question may be trivial, but any insight would be helpful.
0
Jason
12/20/2016 7:47:03 PM
comp.soft-sys.matlab 211264 articles. 26 followers. lunamoonmoon (257) is leader. Post Follow

0 Replies
702 Views

Similar Articles

[PageSpeed] 38

Reply: