Signal processing methods to address some challenges in binaural hearing aids

该提案旨在实现以下长期目标的重要步骤：1)提高助听器的性能和稳健性，并使听力残疾人士更多地接受和使用助听器；2)使助听器以及一般的多声道语音/音频/声学处理系统对困难、具有挑战性的现实生活声学环境，包括各种类型和级别的噪声以及混响条件的竞争说话者的存在，具有很强的耐用性。为了在实现这些目标方面取得进展，提出了三个短期目标。 在双耳助听器(以及其他应用)中，由于不完美的到达方向估计方案、来自助听器用户的小头部移动以及多径传播，在假定用于处理的目标到达方向和实际目标到达方向之间可能存在不匹配或误差。这种失配会导致波束形成算法的性能显著下降，这在今天仍然是一个重要的问题。该建议的第一个短期目标是开发用于双耳助听器的波束形成算法，该算法对估计的目标到达方向的误差具有鲁棒性。 通过无线链路进行耳对耳通信是双耳助听器的另一个挑战，对音频编码的要求具有挑战性，包括低带宽、低延迟(小于1-2ms)、低复杂度和宽带频谱。该提案的第二个短期目标包括开发用于双耳助听器的子带域反向线性预测音频编码方案。我们将加入一个新的噪声整形方案，包括从双耳波束形成算法的白噪声增益中实时反馈，以避免或减少波束形成算法的量化噪声放大。此外，考虑到所使用的反向线性预测编码方案的特定低延迟约束和编解码器同步问题，将为低延迟子带编解码器开发一种定制的分组丢失隐藏方案。 第三个短期目标是进一步发展我们最近提出的一种双耳助听器波束形成方案，该方案既可以降低噪声，又可以保留背景噪声和竞争说话者的双耳线索。我们将研究各种策略，通过利用组件之间的冗余并通过选择允许更好地共享资源的组件来降低方法的复杂性。我们还将研究带宽高效版本算法的开发，在该版本中，需要通过无线链路传输的信号数量最少。这些改进将显著提高实施该方法的助听器的电池寿命。

The proposal aims to achieve significant steps towards the following long-term objectives: 1) To increase performance and robustness of hearing aids and achieve greater acceptance and usage by people with hearing disability, and 2) to make hearing aids and, in general, multichannel speech/audio/acoustics processing systems robust to difficult, challenging real-life acoustic environments, including the presence of competing talkers, of various types and levels of noise, and reverberant conditions. To achieve progress towards these objectives, three short-term objectives are proposed. In binaural hearing aids (and other applications as well), there can be a mismatch or an error between the target direction of arrival assumed for processing and the actual target direction of arrival, because of imperfect direction of arrival estimation schemes, small head movements from the hearing aid user, and multipath propagation. This mismatch can result in significant performance degradation in beamforming algorithms, and it is still an important problem today. The first short-term objective of this proposal is to develop beamforming algorithms for binaural hearing aids which are robust to errors in the estimated target direction of arrival. The ear-to-ear communication through a wireless link is another challenge in binaural hearing aids, with challenging specifications for the audio coding including low bandwidth, low delay (less than 1-2 ms), low complexity, and wideband spectrum. The second short-term objective of the proposal consists of developing a sub-band domain backward linear prediction audio coding scheme for binaural hearing aids. We will add a new noise shaping scheme including real-time feedback from the white noise gain of the binaural beamforming algorithm, to avoid or reduce quantization noise amplification by the beamforming algorithm. Moreover, a custom packet loss concealment scheme will be developed for the low-delay sub-band codec, considering the specific low-delay constraint and coder-decoder synchronization issues of the backward linear prediction coding scheme being used. The third short-term objective concerns the further development of a scheme that we recently proposed for binaural hearing aids beamforming, which can both reduce noise and preserve the binaural cues of the background noise and competing talkers. We will investigate various strategies to reduce the complexity of the method by exploiting redundancies between components and by choosing components that allow better sharing of resources. We will also investigate the development of bandwidth efficient versions of the algorithms, where the number of signals that need to be transmitted through a wireless link is minimized. These improvements will significantly improve the battery life of hearing aids implementing the proposed method.