Theoretical analysis of adaptive noise reduction algorithms for hearing aids

This paper presents the theoretical analysis of 2 different adaptive noise reduction algorithms for twin-microphone hearing aids. A first noise reduction algorithm is based on a beamformer technique [1] and a second is based on optimal filtering and singular value decomposition (SVD) [2]. On the one hand, it is shown that the SVD-based technique gives a robust solution against unmatched microphone characteristics. On the other hand, the beamformer technique has a better robustness against voice activity detector (VAD) errors.