|
k = np.multiply(k, np.sqrt(1.0 / erbs_frequencies)) |
I once had a discussion with @sstrahl about whether this line is correct. Semantically, this solution should be cleaner:
# f_prime / f
q = erbs_frequencies / pitch_candidate
# Apply envelope
k = np.multiply(k, np.sqrt(1.0 / q))
In my tests, I did not see a difference when changing it, most likely because of the normalization step that follows:
|
k = k / np.linalg.norm(k[k > 0]) |
So, maybe this could be changed with future updates?
libf0/libf0/swipe.py
Line 192 in 0baf532
I once had a discussion with @sstrahl about whether this line is correct. Semantically, this solution should be cleaner:
In my tests, I did not see a difference when changing it, most likely because of the normalization step that follows:
libf0/libf0/swipe.py
Line 195 in 0baf532
So, maybe this could be changed with future updates?