仿人工耳蜗处理后频率失匹配及双耳频率互补的声调听感知研究

The current cochlear implant (CI) system fails to provide satisfactory tone or pitch perception ability to the implantees. This deficit could degrade the auditory-speech rehabilitation for those pediatric CI recipients using Mandarin as their mother tongue. The mainstream speech processing strategies of CI nowadays focus on the temporal envelope (TE) of the speech, and introduce more or less spectral cues to these extracted TEs when the speech signals flowed out their corresponding channels. Theoretically, these spectral cues have to be transmitted to the positions responsible for the corresponding frequencies in the basilar membrane so as to optimize the efficacy. However, the stimulating electrodes are not implanted in a absolutely frequency-place matched pattern in most of the time because of the anatomy of the cochlea or some other factors. Additionally, there are inevitable signal interactions between electrodes adjacent with each other due to the limited space of the cochlea. In the present study, the acoustic models—which simulate the monaural or binaural shallow insertion of the electrode array with or without “Zipper Processor”—will be established using acoustical signal processing techniques similar to the working mechanism of CI system. The auditory performance will be recorded based on these acoustic models so as to investigate the impact of frequency-place mismatch on tone recognition as well as the binaural fusion mechanism of auditory system for the split spectral cues. This program is supposed to provide useful cues for the improvement of frequency resolution of CI system. The result data of this study could be the theoretical and experimental basis for the invention of new versions of CI speech processing strategy which would bring increased tone/pitch perception to their users.

人工耳蜗目前难以使植入者获得满意的声调/音调感知能力，这极大影响了以中文普通话为母语的儿童植入者听觉言语康复效果。目前主流的人工耳蜗言语编码策略侧重于对言语信号包络的提取，言语包络通过各自的滤波通道最终被提取出来后即带有相应的频域信息。这些频域信息需要传递至基底膜上相应的频段区域以实现其理想效果。由于耳蜗的解剖学等因素，人工耳蜗电极常较难以实现频率-位置完全匹配性植入。另外，由于耳蜗内的空间局限性，人工耳蜗相邻刺激电极间难以避免信号的相互干扰。本研究拟通过声信号处理技术对人工耳蜗的工作方式进行模拟，构建出单/双侧植入电极浅插入以及双耳间频谱交错嵌合的听感知模型，以研究在频率-位置失匹配条件下听觉系统对声调信息的感知以及双耳间声调信息的整合机制，探索提升人工耳蜗植入者频率分辨率的有效途径，为开发有利于提高植入者声调/音调感知能力的新型人工耳蜗言语编码策略提供全新的理论依据和实验证据。

中文普通话的声调康复一直是听障患者面临的言语康复上的难点，目前大多数的听觉辅助装置包括助听器以及人工耳蜗都无法提供精准的频率分辨率，从而无法给听障患者带来满意的声调/音调感知能力，尽管部分最新的人工耳蜗言语编码策略已能提供较为精准的频率分辨率，然而此类听障患者的听觉神经对电听觉这种新形式所携带信息的提取和利用依然存在问题。另外。由于耳蜗的解剖学等因素，人工耳蜗电极常较难以实现频率-位置完全匹配性植入，这种携带频率和处理频段的不匹配也进一步降低了人工耳蜗植入者感知声调的精确性，从而导致其声调发声的不理想。本项目研究了耳间及耳内的频率不匹配以及双耳间频谱嵌合的听感知模型，并以此听感知模型为基础，进一步研究了人工耳蜗植入儿童在此模型基础上的声调发声能力。另外，对于言语康复中的声调康复，本项目也探索了一种更为客观且高效的声调康复评估方式，这对开发更新一代人工耳蜗言语尤其是声调编码模式，提升人工耳蜗植入者频率分辨率提供了理论及实验依据，并为最终提高听障患者声调康复的有效性提供了切实可行的客观并高效的新方法。.本项目研究发现：（1）由于耳内人工耳蜗携带频率与基底膜处理频段的不匹配，导致先天性感音神经性聋患儿在声调感知的精确性上较正常发育儿童要差，而由于感知-产出之间可能存在的闭合环路，进一步导致了这些患儿声调发声能力明显弱于正常发育儿童。（2）从声学分析的角度，声调信息采集量的多少并不能显著缩小先天性感音神经性聋患儿（植入人工耳蜗后）与正常发育儿童之间的声调能力上的差距。（3）从方法学的角度，神经网络对于声调能力的评估比传统的人耳评估具有相当显著的优势，更为客观且高效。.综上所述，本项目针对人工耳蜗声处理模式对于听觉感知尤其是声调感知的机理及其对声调发声能力的影响进行探索具有重要的科学意义，尤其是对听障患者甚至包括病理状态下的言语障碍患者的声调能力康复具有临床实用意义。