EM, a python package for Gaussian mixtures models

em is a package which makes it possible to create Gaussian Mixture Models (diagonal and full covariance matrices supported), to sample them, and to estimate them from data using Expectation Maximization algorithm in a python (python / numpy / scipy environment)

There is plotting support for confidence intervals/ellipsoids, computation of the so-called Bayesian Information Criterion, and experimental support for recursive EM. Also, em is fast enough to be used in moderate to big size data samples (several tens of dimension, several tens of components, several hundred of thousand of samples).

em is since 2007 a subpackage of the learn scikit: see scikits for more information. The package itself is still usable without learn, though.

AudioLab, a package to do IO with audio files in a python/numpy environment

AudioLab is a python package to import data from audio files into numpy arrays. The goal is to have an equivalent of matlab's audio IO related function like wavread, wavwrite, soundsc and so in the scipy environment, as well as more advanced capabilities. It is just a wrapper around the excellent sndfile library from Erik Castro de Lopo, which gives access to many audio formats, including wav, aiff, htk, ircam and even flac, an open source lossless compressed format. For now, only audio file IO is implemented.

Samplerate, a python package to do high quality samplerate conversion of audio data

samplerate is a python package to convert sampling rate of audio signals; the focus is on quality, by using sinc interpolation and polyphase implementation. For now, samplerate is just a python wrapper around SRC, the Sampling Rate Conversion library, aka Source Rabbit Code, the high quality sampling rate convertion library of Erik Castrop de Lopo, which is based on sinc interpolation.