Assessment of a feasible virtual acoustics method for testing hearing aids using the Hearing-Aid Speech Perception Index

Significant differences have been found in hearing aid (HA) performance between laboratory and real world test environments. Virtual sound environments provide a degree of control and reproducibility which is lacking in real world testing but may require an impractical number of loudspeakers. We assessed the accuracy of a simulation approach in which sources’ direct sound is delivered by single loudspeakers while room acoustics are reproduced using low-order Ambisonics and a small number of loudspeakers. In a large office, we recorded binaural hearing aid output in response to sentence targets and babble noise presented at various levels and from various combinations of four loudspeakers surrounding a manikin. We measured the loudspeakers’ room impulse responses (IRs) using a 32-channel spherical microphone array (Eigenmike), and split the IRs into "direct sound" and "room sound" portions. In an anechoic chamber, the original acoustics were simulated using Ambisonics or discrete loudspeakers for each source’s direct portion and Ambisonics for the room portion. Ambisonic order and/or number of playback loudspeakers were also varied. HA output in the simulations was recorded using the manikin and assessed by comparing Hearing-Aid Speech Perception Index (HASPI) values computed on the simulation recordings with those made in the original room.