Fluid Dynamics and Acoustics Group



Auditory process model (2004-2008)


Psychoacoustic model for sound localisation has been studied in the ISVR for the evaluation of spatial audio systems. Given a pair of signals recorded at (or synthesised for) two ears (binaural signals), the model produces an objective prediction of sound source location, partly based on the known psychological and physiological properties of human hearing system.

1) Peripheral process: acoustic signals arriving at listener's ears are modified in many ways and converted into neural signals in the cochlea, travelling further to different parts of our nervous system including the auditory cortex. From acousto-mechanical measurements and the recordings of neural impulses, transfer functions of the first (peripheral) part of this signal transformation are relatively well known, and can be modelled as a series of linear and non-linear filters.

2) Binaural process: it is known that these monaural neural signals are combined to produce the percept of sound source locations, but detailed neural processes have yet to be fully revealed. We used a recent model of the binaural process suggested by Breebaart and his colleagues, which obtains patterns of excitation-inhibition (EI) cell activities as the internal representation of source location.

3) Central process: these EI patterns contain information of interaural time and level difference, and a simple pattern-matching technique can be applied as a central process to finally provide an estimate of sound source location in the horizontal plane.

Many localisation and lateralisation listening tests have been carried out in the ISVR, and some of the results were successfully compared to the prediction of our binaural hearing model.