Authors: Zi Kei Addison Kwong¹, Ewan Macpherson¹
¹Western University
Background: Measuring and verifying the polar pattern adopted by a hearing aid (HA) with adaptive directionality is challenging because the speech signal and background noise necessary to induce the directional processing will interfere with probe signal measurement. Conversely, probe signals may affect the directionality of the system. This project assessed a novel method of covertly measuring adaptive directionality using simultaneous, inaudible, direction-dependent intensity modulations applied to diffuse background noise.
Methods: Devices to be measured were placed at the center of a circular array of 16 loudspeakers in an anechoic chamber. Each loudspeaker played independent white noise with a unique intensity modulation frequency (MF) between 1000 and 1150 Hz. In some measurements, continuous speech was added from one loudspeaker to induce directional pickup. Coherent averaging of the HA output intensity followed by Fourier analysis at the direction-specific MFs yielded the gains applied to sounds arriving from each loudspeaker direction. The method was applied to a multipattern spherical microphone array, to clinically fit HAs with fixed directionality and typical compression and frequency shaping, and to HAs with a speech-finding beamformer. The effects of different sample lengths, speech to probe noise signal-to-noise ratio, and compression were also investigated.
Results: This method was able to recover the pickup patterns of the multipattern microphone and the fixed and speech-dependent patterns of the hearing aids. With every doubling in recording length, there was a 1.5 dB reduction in the measurement noise floor. As speech to probe-noise SNR increased, so did the noise floor. With speech present at 0 dB SNR, a two-minute recording provided usable results.
Conclusions: This novel method of directionality measurement successfully characterized hearing aid directionality in the presence of speech and dynamic range compression given adequate recording time.
