Measuring the benefit of NELE algorithms for hearing aid users in realistic scenarios with the AFC-MHA platform
When determining the hearing profile of a listener, it is commonplace to measure the SRT (Speech Reception Threshold) with non-reverberant speech against artificially created speech-shaped noise. However, such conditions do not reflect real-world acoustic environments in which speech communication is actually experienced.
In this study we measured the psychometric functions for 20 hearing aid users in two realistic acoustic scenes, which are representative of everyday scenarios: a living room and a busy cafeteria. All the subjects were native German listeners (mean age: 73 years, hearing profiles N3/N4). The German Matrix test was used as speech corpus; binaural recordings of real-world noise and impulse responses were used in order to recreate the acoustic scenarios.
In order to preserve an accurate representation of spatial cues while providing an adequate compensation for hearing loss, we presented the stimuli via headphones using the openMHA  as a simulation of hearing aids. The audio output of the AFC  test platform was routed to the openMHA in real time via the Jack Audio Connection Kit. Stimuli were played back at realistic presentation levels, i.e. 65 dBA for the living room and 75 dBA for the cafeteria, while speech was scaled to match the desired SNR. Given the intensity of the stimuli and the increased loudness sensitivity of the HI (Hearing Impaired) subjects, we used the compressive CR2-NALRP fitting rule, which is based on . Individual SRT50 and slope of the psychometric functions were estimated concurrently with an adaptive procedure.
The results of this study will be used as a starting point for an evaluation of NELE (Near-End Listening Enhancement) algorithms for HI subjects in realistic noise, which follows in the footsteps of a 2019 study with NH native English listeners .
Funding: This project has received funding from the EU’s H2020 research and innovation programme under the MSCA GA 67532 (the ENRICH network: www.enrich-etn.eu).
1. Herzke, Tobias, et al. "Open signal processing software platform for hearing aid research (openMHA)." Proceedings of the Linux Audio Conference. 2017.
2. Ewert, Stephan D. "AFC—A modular framework for running psychoacoustic experiments and computational perception models." Proceedings of the international conference on acoustics AIA-DAGA. 2013.
3. Grimm, Giso, et al. "Implementation and evaluation of an experimental hearing aid dynamic range compressor." Threshold 80.90 (2015): 100.
4. Chermaz, Carol, et al. "Evaluating Near End Listening Enhancement Algorithms in Realistic Environments." Proc. Interspeech 2019 (2019): 1373-1377.