Research Project
9948625
Project Summary Here we conduct research towards new solutions that lead to an improvement and to new types of technical devices for the rehabilitation of hearing impairment. Novel algorithms for dynamic compression, frequency compression, auditory-scene-analysis based noise reduction and speech enhancement algorithms as well as feedback management will be researched, implemented in real-time and evaluated in combination in subjects with different types of hearing impairment. For a realistic assessment of the achievable algorithm benefit, benchmarking and evaluation procedures are researched that involve instrumental and subjective system performance evaluation in virtual acoustic environments of varying complexity that range from simple stationary spatial scenes to complex acoustic scenes with moving sound sources and receiver position. The complete benchmarking system including the different acoustic scenes, the real-time signal processing system including a set of baseline algorithms as well as the baseline instrumental measures will be published open-source as an ?open community platform for hearing aid research?. The platform includes means to develop and integrate own signal-processing algorithms as well as instrumental measures. This way, the platform enables sustainable hearing-aid research and provides the means to do comparative studies based on a quasi- standard set of benchmarks. To foster collaborative research efforts on algorithms, instrumental measures as well as new technical solutions, the project will organize a community challenge based on the open platform. Researchers will be invited to contribute algorithms and/or instrumental measures on the platform, which will then be benchmarked and subjectively tested by the project consortium. Results will be disseminated in a special session of a relevant conference, in addition to publishing the algorithms and results in peer-reviewed journal publications. To foster transfer into applications, the real-time algorithm development system of the platform will also be provided for recent ARM-processor based hardware platforms that allow for the development and evaluation of working demonstrators of wearable audio devices, so-called ?hearables?. Long-term sustainability of the approach is ensured by publishing the open platform including the set of standard processing algorithms and fitting options under an open-source ?copyleft? license, such as the GNU General Public License GPLv3. In addition to this open-source license, a commercial license option allows product development on the basis of the platform in case closed-source components, e.g., specific commercial algorithms, need to be included. This dual license approach is similar to what other academic institutions do, see, e.g., the license options of the FFTW software of MIT. In summary, the project leads to more integrated, sustainable and focused research towards improving hearing devices in general.
