15:40:05	 From  Michael Kirkhart : This might be helpful
15:40:08	 From  Michael Kirkhart : https://stackoverflow.com/questions/2060628/reading-wav-files-in-python
15:40:10	 From  Fredric Harris : in matlab, speedup often happens with preassigning data block lengths... don't know about python 

15:40:21	 From  Leonard Dieguez : Pandas and numpy
15:41:25	 From  Fredric Harris : question: did you apply a window to time series (besides default rectangle window)
15:42:40	 From  Leonard Dieguez : scipy has a signal library for the matlab type of processing
15:42:56	 From  Fredric Harris : I use the derivative to proess time signal till I see an impulse... ten transform doesn't fall.. then integrate back a number of times to match number of dereivatives... spectrum will fall like 1/f^(k+1)
15:43:47	 From  Arthur Lobo : There is also 1/f noise
15:43:54	 From  Radu Pralea : Musical rhythm spectra from Bach to Joplin obey a 1/f power law
https://www.pnas.org/content/pnas/109/10/3716.full.pdf
15:43:55	 From  subbotin : Brewster, not python, but kotlin — there is an open-source project on early stages related to distributed processing of such data (I’m working on it). Check it out if that may help you https://wavebeans.io
15:45:23	 From  Fredric Harris : don't use Gaussian... use Kaiser-bellel
15:45:28	 From  Fredric Harris : bessle
15:46:34	 From  Thad Welch : Back to the convolution idea … convolve 6 pulses and you are within 1% of a Gaussian
15:47:04	 From  Thad Welch : central limit idea
15:47:14	 From  Fredric Harris : it estimates main lobe, but our interest is often tails
15:47:40	 From  Brewster LaMacchia : @subbotin thatnk you for that.  I use sox sometimes.