Developing a portable testing solution for sound localization

AIMS
Sound localization is a fundamental function of hearing. It affects the way a person interacts with the environment. However, the gold standard for sound localization testing, the AB-York Crescent of Sound (CoS)(1), a set of standardised sound sources that are presented to the test subject from reference directions, requires a full or hemi-anechoic chamber for the testing to take place. The footprint of the setup is sizeable, increasing the infrastructural demands of housing such a system, resulting in the CoS only being available in tertiary care centres. There is currently one CoS in active use for diagnostic testing in Scotland.
Driven by developments in the gaming and sound recording industries, algorithms are becoming available to simulate spatialized sound sources using virtual reality. In our work, we aim to create a smaller, portable alternative to the CoS. This could also be used to carry out testing remotely, potentially enabling at-home testing.

METHOD
The experience of using the CoS using a VR headset and a pair of studio/medical-grade headphones is currently being recreated using the Unreal Engine 5 (UE5) virtual reality development environment, coupled to a set of sound spatialization engines, currently UE5 Panning, Steam Audio, Google Resonance, Project Acoustics, Audiokinetic Wwise and FMOD. As part of this, in the present work we investigate the distortions to sound spectrum and attenuation created by these spatialisation engines. In particular, we have simulated pure tone sounds from virtualised sound sources in a realistically sized simulated environment and compared the spectral distortion and attenuation against theoretical values from fundamental acoustics.

RESULTS
Our results so far indicate that the different spatialisation engines present significantly different levels of compliance with well-posed acoustic modelling. We note that, at present, the work is ongoing and, so far, while we note that these non-compliances are indeed present, and well within audible levels, it is yet, unclear if these have a clinically meaningful impact on a person’s ability to localize sound. Therefore, we are currently in the process of completing a participant study that will help us corelate the previously identified spatialization engine characteristics and human sound localization performance. The results of this study will dictate which sound spatialization engine(s) will be taken to the developmental stage of the testing platform.

CONCLUSION
To conclude, our results indicate that, while sound spatialisation engines for virtual environments are available, non-compliances with physical modelling are present, and that these differ between spatialisation engines. To progress towards the development of a portable, more accessible alternative for sound localization testing, we are now planning to develop the testing environment next and carry out a human study to understand the clinical meaning of the non-compliances.

REFERENCES
(1) Kitterick PT, Lovett RE, Goman AM, Summerfield AQ. The AB-York crescent of sound: an apparatus for assessing spatial-listening skills in children and adults. Cochlear Implants Int. 2011 Aug;12(3):164-9. doi: 10.1179/146701011X13049348987832. PMID: 21917204.