AudioSeg For Linux

AudioSeg is a toolkit dedicated to audio segmentation and classification of audio streams. The toolkit implements standard reference algorithms such as energy-based silence detection, BIC segmentation and clustering as well as GMM/HMM classification.

Here are some key features of "AudioSeg":

� silence / audio activity detection � blind segmentation with the Bayesian information criterion � segment clustering � segment classification with Gaussian mixture models � joint segmentation and classification using hidden Markov models

Requirements:

� SPro >= 4.0

Installation follows the standard GNU installation procedure. From the package top level directory, the commands

>> ./configure >> make

will build the library and the runtimes, located in the src directory. Since AudioSeg rely on the SPro 4.0, the location of SPro must be specified with

--with-spro[=path] SPro location (requires SPro 4.0 or greater)

If SPro was installed with the SPHERE package support, this must also be specified with

--with-sphere[=path] SPHERE file format support (requires SPHERE 2.6)

See configure --help for more details on the configuration script options. Installing the library and the runtimes can be done by

>> make install

The installation path is specified by the configuration script (try ./configure --help for details) and defaults to /usr/local.

What's New in This Release:

� switched from SPro 3.3.1 to SPro 4.0 � unified option switches accross tools � unique header and library