DNN-HA: A DNN-based hearing-aid strategy for real-time processing

Release v1.0 of DNN-HA, a deep-neural-network (DNN) hearing-aid (HA) strategy for real-time sound processing. The DNN-HA strategy comprises a single DNN model architecture, and uses an audio signal and an audiogram as inputs to provide individualised HA processing that compensates for the elevated hearing thresholds of an individual listener. This release provides an example script test_DNN-HA_wavfile.py that shows how the DNN-HA model can be used to process an example sentence based on a provided audiogram.

The supporting paper will be presented in ICASSP 2023 – 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), scheduled for 4-9 June 2023 in Rhodes Island, Greece:
"A DNN-based hearing-aid strategy for real-time processing: One size fits all" by Fotios Drakopoulos, Arthur Van Den Broucke, Sarah Verhulst