Development and evaluation of psychoacoustically motivated binaural noise reduction and cue preservation techniques

Due to their decreased ability to understand speech hearing impaired may have difficulties to interact in social groups, especially when several people are talking simultaneously. Fortunately, in the last decades hearing aids have evolved from simple sound amplifiers to modern digital devices with complex functionalities including noise reduction algorithms, which are crucial to improve speech understanding in background noise for hearing-impaired persons. Since many hearing aid users are fitted with two hearing aids, so-called binaural hearing aids have been developed, which exchange data and signals through a wireless link such that the processing in both hearing aids can be synchronized. In addition to reducing noise and limiting speech distortion, another important objective of noise reduction algorithms in binaural hearing aids is the preservation of the listener?s impression of the acoustical scene, in order to exploit the binaural hearing advantage and to avoid confusions due to a mismatch between the acoustical and the visual information. This can be achieved by preserving the binaural cues, i.e. the Interaural Level Difference (ILD the Interaural Time Difference (ITD) and the Interaural Coherence (IC) of all sound sources in the acoustical scene. Considering the importance of the binaural cues for speech intelligibility and spatial awareness, the main objective of this thesis is to develop and evaluate algorithms for noise reduction in binaural hearing aids, which, in addition to preserving the binaural cues of the speech component, also preserve the binaural cues of the noise component. Generally, the proposed algorithms are based on the binaural multi-channel Wiener filter (MWF), since this technique advantageously combines spatial filtering with spectral filtering and preserves the binaural cues of the speech source. We propose several extensions of the binaural MWF aiming to preserve the binaural cues for acoustic scenarios with 1) a single desired speech source in a diffuse noise field, 2) a single desired speech source with an additional interfering source in a diffuse noise field. For acoustic scenarios with a diffuse noise field, we propose an extension of the binaural MWF, namely the MWF-IC, aiming to also preserve the IC of the diffuse noise field. Since for the MWF-IC no closed-form solution exists, we also propose to preserve the IC of a diffuse noise field using both the binaural MWF with partial noise estimation (MWF-N) and the binaural MVDR beamformer with partial noise estimation (MVDR-N), for which closed-form solutions exist. Since for all proposed algorithms a trade-off between IC preservation and noise reduction performance exists, depending on trade-off parameters, we propose to select these trade-off parameters based on the IC discrimination ability of the human auditory system, such that an optimal trade-off between noise reduction performance and spatial awareness preservation is obtained. In addition, for the MVDR-N beamformer we derive a closed-form expression for the trade-off parameter yielding a desired Magnitude Squared Coherence (MSC) for the output noise component. Furthermore, we evaluate the proposed algorithms using objective measures and subjective listening tests, showing that the proposed algorithms always improve the spatial impression of the output signal and can in some cases even increase speech intelligibility compared to the binaural MWF. Since previously proposed extensions of the binaural MWF are not able to achieve perfect binaural cue preservation for both the speech source and a directional interfering source, for acoustic scenarios with an interfering source, we propose two extensions of the binaural MWF. The first extension, denoted as MWF-RTF, aims to preserve the binaural cues of the interfering source by adding an RTF preservation constraint to the binaural MWF cost function. The second extension, denoted as MWF-IR, aims to completely suppress the interfering source by adding an interference rejection constraint to the binaural MWF cost function. Since for both extensions the impact of these additional constraints on speech distortion, noise reduction and binaural cue preservation performance will be different, we provide a rigorous theoretical analysis and comparison of the performance of the binaural MWF, MWF-RTF and MWF-IR algorithms. The theoretical analysis is validated by simulations using measured Acoustic Transfer Functions of a binaural hearing aid in a reverberant room, showing that the performance of the binaural MWF, MWF-RTF and MWF-IR highly depends on the position of the interfering source and the number of microphones. Furthermore, simulation results show that the MWF-RTF achieves a very similar overall noise reduction performance as the binaural MWF, while preserving the binaural cues of both the speech source and the interfering source, whereas the overall noise reduction performance of the MWF-IR is significantly degraded compared to the binaural MWF and the MWF-RTF. In addition, we mathematically analyse the relations of the MWF-RTF and the MWF-IR to the recently proposed BLCMV beamformer, showing that for a special case of the MWF-RTF, the MWF-RTF is equal to the BLCMV beamformer with the interference rejection parameter maximizing the signal-to-noise-plus-interference ratio. Furthermore, we will show that the MWF-IR can be decomposed into a special case of the BLCMV beamformer and a single-channel Wiener postfilter.