双耳数字助听系统声源快速分离及声信号完全可重构分解研究

In this project, two key problems in improving the intelligibility of speech for binaural digital hearing aid system will be studied. The aim of theses studies is to propose innovative approaches to enhance the performance of digital hearing aid system in complicated environments. (1) To solve the problems of fast sound source extraction and binaural cues retention, mixed-signal separation algorithm based on the combination of frequency bin selection and frame selection will be proposed. In addition, spatial information is added to the cost function for a joint optimization. (2) To enable the digital hearing aid system to optimize its sound processing strategy in accordance with patients’ pathological conditions, completely reconfigurable sound wave decomposition filter bank will be proposed based on fractional interpolation filters and nonlinear mapping. The research will provide new theoretical basis and implementation methods for the development of binaural digital hearing aid system and provide the necessary technical support for the independent development of digital hearing aids in our country.

本课题研究双耳数字助听系统提升语音可懂度的两个关键问题，旨在提出创新性的方法以提升数字助听系统在复杂环境中的表现。（1）构建双耳距离声源分离模型，提出频域频点筛选和时域帧筛选相结合的混合信号分离算法，并在此过程中对代价函数加入含有空间信息的约束条件进行联合优化，解决双耳数字系统复杂环境中声源快速提取的问题。（2）提出基于分数内插滤波器和非线性映射的数字助听系统完全可重构声信号分解滤波器组新型结构及实现算法，解决数字助听系统基于听损患者自身病理情况进行声信号分解策略优化的问题。本课题的研究将为双耳数字助听系统的发展提供新的理论依据与实现方法，为我国自主进行新型双耳数字助听设备的开发提供必要的技术支撑。

本课题研究双耳数字助听系统提升语音可懂度的两个关键问题。（1）在盲信号分离快速算法方面取得了创新性成果。通过对混合信号特性的数学分析，对各个频点上的分离性能进行估计并据此进行频点筛选，大大减小了计算复杂度，使盲信号分离算法在实时性要求高场景中得以应用。（2）提出了基于耳蜗特性的低功耗非均匀数字滤波器组及模块化完全可重构滤波器组设计方法，解决了数字助听系统基于听损患者自身病理情况进行声信号分解策略优化的问题，与当前广泛采用的均匀滤波器组相比取得了更优的听力补偿性能。