Abstract
This paper presents a new training system for 3D audio space localization, part of a more complex project, Sound of Vision, which is intended to be an object recognition and navigation aid system for visually-impaired people.
As a large number of people are suffering from a visual handicap which impedes them from normally accomplishing their daily chores, there is need for an assistive device that replaces sight with another sense, for instance, hearing. The most important aspect in the sound localization techniques is the human capacity to discriminate between different sounds playing at various locations in space. The aim of the presented audio system is to train and test the subjects' space localization ability, hence to develop and improve their 3D recognition skills.
The solution uses binaural sounds conveyed through headphones as basic blocks, applying HRTFs to various sounds (Head Related Transfer Function - a physical propagation characteristic that describes how the ear perceives the sound coming from a particular point in space).
The acoustic signals represent short (e.g. 20ms long) white and pink noise sounds, externalized with non-individualized HRTFs recorded in dedicated lab environments. Thus, from a monaural noise sound, through the process of mathematical convolution with the corresponding angular HRTF impulse for both the left and right channels, we obtained a 3D binaural sound which is perceived as coming from the particular direction specified by the angular value of the convoluted HRTF.
The program is composed of a training module, where the user moves the mouse cursor inside a circle and hears a continuous binaural sound that corresponds to the azimuth angle indicated on the circle- ranging from 0 to 355 degrees, and a test module- where the subject is presented 10 different sounds that correspond to 10 different locations, being required to point out on the circle the presumed emitting location of the sound, as he perceives it. Through learning and training, the users (both normal-sighted and visually impaired subjects) performed various auditory tasks- identification, recognition, discrimination- and improved their sound localization capacity.
This paper gives on overview of 3 aspects: the technical realization of the system, the experimental results with a set of training subjects and the advanced observations about sound localization that we made. |