SigProc is a python package and command line tool for facilitating creation of features from audio files. SigProc can be used to generate MFCC, SDC, and Chromagram features. It supports opening wav and HTK PLH file formats, with ffMpeg or Sox any audio file can be opened.
This tool is written for learning, not performance. However it is very quick to generate features, even on large data sets. It is also an experiment/exercise in self documenting code.
SigProc operates on "recipes" which are composed of several building block "processes". Processes are connected in a Pipeline chain. Inputs are automatically connected to outputs of a previous process by the Pipeline Process Runner.
The primary inter-process data types are 'Signal', 'Matrix', and 'Track'. a Signal is 1 dimensional representation of a (mono) audio signal. A Matrix is a 2 dimensional representation, for example a spectrogram. Finally, a 'Track' contains sparse information associated with time. All formats keep track of metadata that was used to generate the object. For example a representation of a spectrogram keeps track of the parameters used, so that the frequency bins can be computed.
Python 3.5 or greater ( may work for 2.7, 3.4 )
requires : numpy scipy optional : PIL (pillow), matplotlib
python setup.py installTo generate documentation for all processes, execute:
sigproc --doc > sigproc.txtTo generate an image representation of the spectrogram at 100 frames per second, execute:
sigproc --dir=./recipes spectrogram16k100 in.mp3 out.pngTo generate an image representation of MFCC features, execute:
sigproc --dir=./recipes mfcc_16k in.mp3 out.pngExample for using SigProc as a python package. Demonstrates loading recipes, generating the pipeline, and running the pipeline on a single input file.
import SigProc
inFile = "in.mp3"
recipe_directory = "./recipes"
recipe = "spectrogram16k100"
# load the specified recipe
rm = SigProc.newRecipeManager(recipe_directory);
procs = rm.getRecipe(recipe)
# create a process runner for the pipeline
pipe = SigProc.PipelineProcessRunner(procs);
# execute, returns instance of SigProc.SpectralMatrix()
result = pipe.run(inFile)[-1]