This code has two main scripts: train_codebook.py, which trains a VQ codebook on a directory of audio files; and sig2hac.py, which converts a single audiofile to its HAC representation.

To train a codebook on directory /path/to/audiofiles/ and store the output in codebook.pkl:

python train_codebook.py /path/to/audiofiles/ codebook.pkl

Note that the script will recursively search the input directory for files with the .wav extension.

To convert audiofile myfile.wav to its hac representation myhac.pkl using codebook mycodebook.pkl:

python sig2hac.py mycodebook.pkl myfile.wav myhac.pkl

• python 2.7.x
• numpy
• scipy
• scikits.audiolab
• scikits.samplerate
• sklearn