Preserving Auditory Cues for Human Echolocation Training: A Geometrical Acoustics Study Using a Benchmark Dataset (BRAS)

Abstract

Human echolocation is a method mainly used within the blind community to navigate using sound emission and analysing the returning echoes from the surrounding environment. Echolocation is predominantly trained by orientation and mobility instructors at visual rehabilitation centres. However, systematic guidelines or protocols focusing on the requirements of room acoustic simulations to accurately represent the auditory cues necessary in a virtual training environment for echolocation have not yet been developed. This paper sets out to investigate the use of geometrical acoustic (GA) calculations for a virtual echolocation training system comparing the measurements from the Benchmark for Room Acoustical Simulation (BRAS) dataset with other GA calculations outcomes. Three simple and one complex test scenes are chosen from the dataset. The calculation settings are optimised for each test scene considering the complexity of the scene, room volume and acoustic phenomena. The monaural room impulse responses from the simulations are analysed with respect to the timing of the reflections and the level relations between the reflections and the resulting frequency response for each scene. These are subsequently compared with each measured counterpart. The paper discusses the results, their limitations, and provides recommendations on the use of GA calculation tools for echolocation training scopes.