Models of binaural hearing for sound lateralisation and localisation
University of Southampton, Institute of Sound and Vibration Research,
The current study suggests two models of binaural hearing, which aim to make predictions for inside- and outside-head localisation of a single sound source in the horizontal plane. Both models consider free-field ITDs and ILDs as the memory of sound localisation to which the target interaural disparity is compared. The first model, the characteristic-curve (CC) model acquires the best estimate of a source location by finding the nearest-neighbour of the target ITD and ILD in the characteristic curve of free-field interaural disparities. On the other hand, the second model, the pattern-matching (PM) model, assumes that the excitation-inhibition cell activity pattern suggested by Breebaart et al. [J. Acoust. S. Am., 110(2):1074-1088, 2001] provides the internal representation of the sound localisation cues. Given the uniqueness of EI-patterns, the pattern-matching process operates in each auditory frequency band to give an estimate of the sound source position, which is then frequency-weighted to finally establish the probability function of target location. In the two listening tests presented in the current study, it has been found that both models are capable of predicting many important features of human sound localisation. For example, the inside-head localisation (laterality) of dichotic pure tones has been reasonably well predicted at low source frequencies, 600 Hz and 1200 Hz, by the CC model individualised for each participant. In addition, the prediction of the PM model has been successfully compared to listening test results where the outside-head localisation of the participants was investigated for real and virtual acoustic sources. Given the simplicity and the originality in modelling the central processes of auditory spatial hearing, particularly in handling the ILD information of binaural signals, the predictive scope of the models is regarded as being worthy of further investigation. Furthermore, considering the reasonable predictions made for both lateralisation and localisation of acoustic stimuli, the models developed appear also to be well-suited to the computational evaluation of spatial audio systems.
||University of Southampton
||18 Aug 2010 13:10
||18 Apr 2017 03:46
|Further Information:||Google Scholar|
Actions (login required)