openMHA 4.5.4 and 4.5.5 released
In release 4.5.4 openMHA is extended with a set of plugins that implement a probabilistic sound source localization algorithm. The method applies machine learning techniques and provides the probability of sound source presence for a given set of directions the underlying model has been trained on.
A live example that demonstrates automatically controlled spatial filtering with a 4-channel binaural hearing aid setup is provided: The beamformer, which was previously available in openMHA, is steered based on sound source probability. A web-browser based visualization tool of the localizer’s output is also included. This demonstrator was developed at University of Oldenburg in the Research Unit FOR1732 “Individualized Hearing Acoustics”. Furthermore the set of plugins for signal handling was extended - amongst others an all-purpose resampling plugin providing enhanced resampling functionality and quality was added.
Changes in release 4.5.5 are related to a fix in the calibration settings of the mhacontrol tool for Octave/Matlab following user requests (Github issue #4).
See Download for details.