Welcome to the source page for my University of Miami MMI 505 project. With millions of sound recordings now available at the click of a mouse, the ability to find music in useful ways that go beyond a traditional textual metadata search, is becoming increasingly important and viable. My project was to write a Python module that extracts low-level acoustic features from an audio file (.wav or .mp3) on a frame-by-frame basis and computes meta-features based on these vectors.
Use like this:
from acousticfeatures import * x = FeatureExtractor('White_Noise.wav') #can read .wav or .mp3 x.process_frames() feature_dict = x.return_features()
feature_dict now contains a dictionary with entries for each of the features. The implemented features are: Mean and Standard deviation for these time vectors: RMS, Spectrum Centroid, Spectrum Spread, Spectrum Skewness, Spectrum Kurtosis and Zero Crossings.
$ hg clone http://bitbucket.org/reiddraper/acousticfeatures/wiki/