July 26, 2010 in MATLAB by Hagen Wierstorf | No comments

A new version of the growing Auditory Modelling Toolbox has been released. The main new feature is a binaural model after Lindemann (1986a) that uses a running cross-correlation with inhibition to predict the perceived lateralization of an auditory event. The output of the model depends on the auditory filters and on the time (see figure below).

To install it, you can download it from Sourceforge. You also have to install the Linear Time/frequency Toolbox. Then in Matlab or Octave just go to the directories of the toolboxes and run:

>> ltfatstart
>> amtstart

After this you are able to use the Lindemann model, see help lindemann for an introduction. To produce the figure below, you can run:

>> demo_lindemann;

Another very nice feature is the function exp_lindemann1986a that is able to reproduce the figures of the Lindemann paper. For example to reproduce figure 6 of the paper, just type exp_lindemann('fig6'). The model is under further development and will include in the next release a version of the method proposed byGaik (1993) to identify natural combinations of ITDs and ILDs. Further a version of theBreebaart model will also be included in one of the next releases.

The paper is available in the online proceedings.

Below is a screenshot of the interface to the patch.  The patch set itself is presently available as a beta by request.

AmbiGrainer is a high quality granulation engine, with sub-block grain timings, capable of producing some very smooth sounds. Grains may be individually panned in 3rd order Ambisonics (with per-grain order specification). Grains may also be specified directly by an external algorithm, or at a meta-level using the provided GUI. A nice feature is the creation of an Ambisonic grain cloud of grains at randomised directions.  The system can run many thousands of grains per second, with 80 simultaneous voices running easily on a 2007 MacBook Pro.