Kristin Ohlmann1, Florian Denk,2 Birger Kollmeier1
1Medical Physics, University of Oldenburg, Oldenburg, Germany
2German Institute of Hearing Aids, Lübeck, Germany
Acoustic transparency – i.e., the ability of a hearing device to “just amplify without side effects” by providing an overall transfer function to the eardrum mimicking the open ear – is a desirable feature to increase the spontaneous acceptance. However, an individual equalization filter accounting for individual ear acoustics is difficult to achieve in real hearing devices, because the required individualized acoustic transfer functions are not available for most applications unless measurements at the eardrum are performed. Furthermore, the processing latency and the leakage of external sound into the ear canal can deteriorate the perceived sound quality. We therefore assessed how sound quality is influenced by different parameters, such as using individually measured or generic data from a dummy head to compute the equalization filter, as well as different latencies or vent configurations. Also, we evaluated how a direction-independent approximation of the direction-dependent transfer function from hearing device microphone to eardrum can be obtained, which allows for a good estimation of the signal at the open ear. Based on individually measurement data, a hearing device was simulated in 10 normal hearing participants’ ears using individual binaural synthesis, allowing to freely vary all hearing device parameters. Acoustic scenes “heard” through the different virtual hearing devices were presented in a MUSHRA-like framework via headphones, and the overall sound quality was rated by the participants. Quality ratings for conditions with individualized transfer functions tend to be higher than for generic data, while latency and leakage show no large overall influence. Differences tend to be larger for complex scenes. In addition to the subjective listening tests, objective quality measures were obtained through a set of models. A good agreement between subjective and objective data was achieved using the GPSMq model. This will support the interpretation and parameter optimization for acoustic transparency in the future.