README.md

MDCTN 🧶

Multidimensional Modified Discrete Cosine Transforms

pip install mdctn

• 1-D MDCT & IMDCT
• n-D MDCT & IMDCT
• Windowing support
• Helper functions for signals

1-D MDCT on signals

Signals are wrapped around so all the data is stored in the same number of bits.

import numpy as np
from mdctn import mdct, imdct

x = np.arange(24)

y =  mdct(x, N=16)
z = imdct(y, N=16)

np.allclose(x, z) # True

1-D Pure MDCT

The core MDCT function

import numpy as np
from mdctn import core

x = np.arange(6) # [0, 1, 2, 3, 4, 5]

y_1 = core.mdct(x[0:4]) # [-2.50104055, -0.49476881]
y_2 = core.mdct(x[2:6]) # [-4.34879961, -1.26013568]

z_1 = core.imdct(y_1) # [-0.5,  0.5,  2.5,  2.5]
z_2 = core.imdct(y_2) #             [-0.5,  0.5,  4.5,  4.5]

z = (z_1[2:4] + z_2[0:2]) # [2.0, 3.0]

Benchmarks

See benchmarks.ipynb