gms | German Medical Science

Research question: The use of virtual acoustics provides the opportunity to evaluate hearing aids in complex, but controllable environments that are more realistic than standard laboratory tests. The experimental setup, however, requires typically 16 or more loudspeakers. In order to reduce the setup requirements and make it more accessible to clinics, this study investigated whether binaural synthesis of virtual complex acoustic environments is suitable for speech intelligibility testing in the laboratory. We investigated the extent to which data obtained with a binaural synthesis of hearing aid microphone signals are comparable to those obtained with a loudspeaker array.

Methods: The input signals of a set of two-channel binaural behind-the-ear (BTE) ear-level devices were generated from impulse responses measured in a laboratory using a 16-loudspeaker array. The 80% speech reception threshold (SRT80) was determined for 20 hearing-impaired listeners using the German Matrix Sentence Test (OLSA, [1]). The listeners were provided with a research hearing aid based on the open Master Hearing Aid (openMHA, [2]) software and with realistic BTE hearing aids, which were fitted with individual earmolds. Gains were fitted according to a loudness-based fitting rule. In addition, a monaural and a binaural signal enhancement strategy was used: Bilateral adaptive differential microphones and a binaural minimum variance distortionless response beamformer. The listening tests were conducted in three different complex virtual acoustic environments rendered with the Toolbox for Acoustic Scene Creation and Rendering (TASCAR, [[3]): two scenes with localised speech-like interferers (symmetric/asymmetric) and one with a diffuse background noise (Cafeteria).

Results: SRT80s between loudspeaker presentation and binaural synthesis were highly correlated with an R² between 0.59 and 0.84. The intra-individual standard deviation for the method under test was 0.47 dB to 1.06 dB higher than the test-retest reliability of the reference method of 1.14 dB to 1.84 dB.

Conclusions: The results indicate that binaural synthesis of hearing aid input signals is suitable for performing speech intelligibility tests in complex acoustic scenes for the evaluation of hearing aid algorithms.

Work funded by the DFG – EXC 2177/2 - Project ID 390895286.