Dissertation

1. Summary

This thesis introduces a new method for research in auditory localization and presents its application to the fields of real and virtual acoustics, auditory-visual interaction, and audiology.

A new localization method was developed, which utilizes a laser pointer to display the localized direction. By turning the ball on a trackball the subject moves the spot of the laser pointer on a horizontal track to the perceived auditory direction. The initial trial-by-trial position of the laser spot is varied symmetrically around the presented direction of the sound in a random manner. This variation reduces methodical bias effects, as the arc travelled by the laser spot does not depend on the direction of the sound. The indirect pointing with the trackball causes a decoupling from the human proprioceptive directional system. Therefore the method is named ProDePo - Proprioception Decoupled Pointer. By using the trackball the new method can be used fast and intuitively, which allows for a wide range of applications. The reduced methodical influence on localized directions and the proprioceptive decoupling make it a unique pointer method which is suitable for experiments in the field of auditory-visual interaction. Localization experiments using the new method show a small overestimation of the sound direction, but less localization error and variance than most alternative localization methods, especially at lateral angles. Using a fixed frontal initial position of the laser spot a small underestimation of sound direction occurs and the variance decreases further, which is ascribed to the introduction of a visual frame of reference for auditory localization.

The new localization method is used to investigate localization in virtual acoustics. To synthesize virtual acoustics, auditory localization cues are reproduced over headphones using head related transfer functions (HRTFs). To select an individually optimized pair of HRTFs from a catalogue of non-individual HRTFs a two step procedure is introduced. In the first step 5 HRTFs are selected from the catalogue, in order to maximize the spatial perception in the frontal area for horizontally moving sources. From these 5 HRTFs a single one is selected which reproduces frontal and horizontal directions exactly, at a non-elevated position, at constant distance, and outside the head. The selection procedure for non-individual HRTFs is evaluated in localization tests. The results show that the selection minimizes the localization variance and error, and the number of inside-the-head localizations, i.e. usually a HRTF is found which supports an externalized, focused image. Localization experiments with selected HRTFs show, nevertheless, a small increase in variance and error in the horizontal plane compared to individual HRTFs, which in turn reproduce horizontal directions comparable to the free field.

The ProDePo-method was further used to study the ventriloquism effect. This describes a bias on auditory directions towards a concurrently emerging visual target. The results of the study show similar bias effects for the auditory direction in the free field and the virtual acoustical environment using individual HRTFs. Using selected non-individual HRTFs the auditory directional displacement towards the visual target is reduced. This contradicts the hypothesis that a reduction of individual information in auditory directional cues renders the auditory direction more susceptible to visual bias. Further investigation shows that the increase in localization variance occurring with non-individual HRTFs is not the cause for the decrease in auditory-visual bias. The relative bias effects are shown to decrease exponentially with increasing distance between the auditory and the visual target, whereas the reduction of bias between the hearing environments can be described by a constant factor.

The intuitive and fast handling of the new localization method allows clinical studies of patients with impaired hearing. Localization experiments on subjects with cochlea implant (CI) using a hearing aid (HA) or a second implant on the contralateral side show unexpected localization ability in both groups. All bilaterally implanted subjects and three subjects of the CI+HA group are able to localize. Of the CI+HA group a further four subjects can discriminate right and left, whereas half of the subjects of this group show no directional hearing ability. A further study with a bilaterally implanted subject shows clearly that interaural level cues are used for localization with bilateral CI. It is also proved that interaural temporal cues are not evaluated at low frequencies or through the envelope using bilateral CI.

The successful application of the ProDePo-method in several different fields -- research, virtual acoustics, and audiology -- verifies the intuitive concept and the versatile applicability of the new localization method.

2. Institute

Prof. Hugo Fastl supervised the thesis. It was conducted at the chair for Mensch-Maschine Kommunikation (Human-Machine-Communication) in the Department for Electrical Engineering and Information Technology at the Technische Universität München.

3. Download

The official publication of the thesis can be found online at the library of the TU München.
The thesis can also be downloaded as a pdf-document here (1.4 MB, in German). Adobe Acrobat Reader will be needed for viewing.