I have a numpy array of a certain length, and want to produce a much shorter array, where each value in the shorter array is aggregated based on a bin of values in the original array. In my specific case, this happens to be an audio signal and the function I’m applying is a max over each bin.

It feels like there must be a function like numpy.interp or scipy.interpolate.interp1d which could make this easier. I swear there must be a simpler/cleaner way to do this with a numpy or scipy built-in.

This is the code I write now:

# assume a is an np.array with len(a) == 510
shorten_length = 50
a_shortened = np.zeros(len(a) // shorten_length)

for i in range(len(a_shortened)):
    slice_end = min(len(a), (i+1) * shorten_length) - 1
    _bin_max = np.max(a[i * shorten_length:slice_end])
    a_shortened[i] = _bin_max