Introduction
These opcodes present variations on the following transforms: Haar, Walsh/Hadamard Cosine
At current the opcodes will function simply and similar to the fft and fftinv opcodes in Csound - ie, the user must implement windowing and ksmps in respect to the desired analysis.
These are experimental and exact usage/manipulation of the functions is subject to some testing so there may be further development to make the opcodes more usable.
Tested on Linux with Csound 6.14
Installation
mkdir build && cd build
cmake ..
make
sudo make install
Opcodes provided
Each of the following opcodes receives and emits a k-rate array. The size of the input array should be a power of two. The output size will match the input.
tfcosine
Cosine transform. Notably slow (probably not for realtime usage)
tfcosineinv
Inverse cosine transform. As Above
tfhaar1
Haar transform, algorithm 1
tfhaar1inv
Inverse Haar transform, algorithm 1
tfhaar2
Haar transform, algorithm 2
tfhaar2inv
Inverse Haar transform, algorithm 1
tfwalsh1
Walsh/Hadamard transform, algorithm 1
tfwalsh2
Walsh/Hadamard transform, algorithm 2
tfwalsh2inv
Inverse Walsh/Hadamard transform, algorithm 2
Credits
Code, inspiration and relevant licensing has been derived from the following: https://people.sc.fsu.edu/~jburkardt/f_src/walsh/walsh.html https://people.sc.fsu.edu/~jburkardt/f_src/haar/haar.html https://people.sc.fsu.edu/~jburkardt/f_src/cosine_transform/cosine_transform.html https://github.com/mochow13/competitive-programming-library/blob/master/Math/Fast%20Walsh-Hadamard%20Transform.cpp
Todo / possibly forthcoming Implement Daubechies wavelets, sine transform.