SPEECH PATTERN RECOGNITION IN ELECTRIC HEARING

电听觉中的语音模式识别

• 批准号: 6176942
• 负责人: Qian-Jie Fu
• 金额: $ 5.42万
• 依托单位: HOUSE RESEARCH INSTITUTE
• 依托单位国家: 美国
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-08-01 至 2002-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6176942
• 关键词: auditory stimulus; behavioral /social science research tag; clinical research; cochlear implants; human subject; medical implant science; medical rehabilitation related tag; neural information processing; psychoacoustics; sound frequency; speech recognition

Although multichannel cochlear implants have achieved great success, there is a wide range of performance across patients. Understanding the cause of high variability of speech performance in cochlear implant users is still one of the biggest challenges in electric hearing. The present studies are aimed at understanding the factors that contribute to individual performance and at improving performance through modifications to the speech processor parameters. The hypothesis of this research is that speech perception in electric hearing is primarily based on a measure of similarity between the electrically-evoked peripheral neural discharge patterns and the central speech pattern templates. We further hypothesize that the central speech pattern templates can accommodate to some degree based on experience with new peripheral neural patterns. Two fundamental aspects of speech pattern recognition in electric hearing are addressed in this proposal. The aim of the first experiment will be investigate how the peripheral neural patterns are affected by speech processor parameters. The match between the peripheral neural patterns and the central pattern templates will be inferred from speech performance. Recognition of vowels and consonants will be measured without any training period. The interactive efforts of three critical speech processor parameters: number of electrodes, the allocation of frequency regions to electrodes, and the stimulation mode and location of electrodes, will be investigated. The second experiment is aimed at understanding how the central speech pattern templates are altered by experience with new peripheral neural patterns of speech sounds. Experimental speech processors with significantly different peripheral neural patterns of speech sounds will be implemented in the subject's wearable speech processor. Speech performance as a function of the same speech processor parameters as those in Experiment 1 will be measured over time. Results from this study may indicate how far the central speech pattern templates can be altered and how long it takes. We believe that the results obtained from this proposal could improve speech performance of individual cochlear implant users through modifications to the speech processor parameters, lead to better designs of auditory prostheses, and improve understanding of speech pattern recognition in both acoustic and electric hearing.

尽管多通道人工耳蜗已经取得了巨大成功，但 患者的表现多种多样。了解原因 人工耳蜗使用者的言语表现的高度变异性仍然存在 电听力的最大挑战之一。目前的研究是 旨在了解影响个人的因素 性能并通过修改来提高性能 语音处理器参数。这项研究的假设是 电听觉中的言语感知主要基于对 电诱发周围神经放电之间的相似性 模式和中心语音模式模板。我们进一步假设 中心语音模式模板可以在某种程度上适应 基于新的周围神经模式的经验。两个基本 电听力中语音模式识别的各个方面在 这个建议。第一个实验的目的是研究如何 周围神经模式受语音处理器参数的影响。 周围神经模式与中枢神经模式的匹配 模板将从语音表现中推断出来。元音识别 辅音的测量无需经过任何训练。这 交互努力的三个关键语音处理器参数：数量 电极的数量、电极频率区域的分配以及 将研究刺激模式和电极位置。这 第二个实验旨在了解中心言语模式如何 模板会根据新的周围神经模式的经验而改变 说话的声音。实验语音处理器具有显着不同 语音的周围神经模式将在 对象的可穿戴语音处理器。 语音表现作为函数 与实验 1 中相同的语音处理器参数将是 随着时间的推移测量。这项研究的结果可能表明 central speech pattern templates can be altered and how long it takes.我们 相信从该提案中获得的结果可以改善语音 通过修改个体人工耳蜗使用者的表现 语音处理器参数，导致更好的听觉设计 假肢，并提高对语音模式识别的理解 声学和电学听力。