Holger
Holger
I found the cause of the error. It can be easily fixed. The precision of pywt.integrate_wavelet must be increased _cwt.py: lines 77-83 precision = 10 int_psi, x = integrate_wavelet(wavelet, precision=precision)...
precision = 10 int_psi, x = integrate_wavelet(wavelet, precision=precision) step = x[1] - x[0] j = np.floor( np.arange(scales[i] * (x[-1] - x[0]) + 1) / (scales[i] * step)) if np.max(j) >=...
I investigated the problem in more detail. It cannot be easily fixed, higher precisions does not solve the problem. At higher scales the convolution shows artifacts: In the following, the...
If the convolution is used with mode='valid' instead of full, the result is the following: ![figure_1-5](https://cloud.githubusercontent.com/assets/2202263/25122602/cea1bb24-2425-11e7-907d-ec8340711d33.png)
We can add the mode parameter to cwt. mode = 'valid' means then, that only the valid part is shown and the rest is zero. mode = 'full' means that...
https://github.com/rafat/wavelib/commit/318e9b0f86d76e3c7e7fd428f7b44cff592ff5b2 Here are 102 coeffs available. But no reference given, from where the coefficients are taken.
In this paper, it is shown that 66 FIR coefficients are sufficient: https://ccrma.stanford.edu/~dattorro/Wavelet.pdf But, no coefficients are listed. Only the method how to compute them...
Try conda install steem :) https://steemit.com/utopian-io/@holger80/install-steem-python-easily-by-conda-forge
PyCrypto is broken and it is not possible to compile it on windows without patching it first. Patches can be found in the conda feedstock for example.
Try conda install steem