A STUDY ON A NEW COCHLEAR IMPLANT SYSTEM BASED ON A PERIPHERAL AUDITORY SYSTEM MODEL

周边听觉系统的数值模型被倒置。这是通过解决Meddis的内毛细胞模型的逆问题，使用在耳蜗中具有非线性特性的转导量来实现的。在这个项目期间，三种类型的听觉滤波器，即Ohdaira的FIR滤波器组，Patterson的伽玛音滤波器组，和Sneff的IIR滤波器组，作为一个逆滤波器的幅度和相位特性进行了检查。我们发现，Ohdaira的过滤器优于其他两个过滤器。通过对前向听觉模型处理后的听觉放电模式进行反向处理，可以得到高质量的再现语音。用MATLAB和Mathematica编制了相应的程序。本文的研究成果将在今年夏天的TCPhS’99会议上发表。我们在一种高效的语音编码技术的基础上，改进了一种多通道人工耳蜗的语音信号到电极脉冲的转换算法。这个问题被分析为一个压缩的语音信息，其中听觉滤波器一个接一个地切换。该过程被简化为找到双极脉冲的一个接一个刺激序列与给定滤波器组的最佳组合。编码质量的感知评价与编码语音的再合成。相对低于低频分量的高频分量没有被很好地编码，并且几乎听不见。这一结果是在悉尼举行的澳大利亚ICSLP'98国际会议上提出的。我们可以与澳大利亚先进的人工耳蜗研究人员进行有趣的讨论，其中包括多通道人工耳蜗系统的著名发明者G.M.Clark。他们对我们的做法印象深刻，向我提出了许多问题，并建议进一步合作。在本项目期间，我们的人工耳蜗研究得到了国际认可，并对人工耳蜗研究史做出了重要贡献。

A numerical model of the peripheral auditory system was inverted. This is realized by solving an inverse problem of the Meddis's inner hair cell model using the amount of transduction with nonlinear characteristics in the cochlea. During this project term, three types of the auditory filters, that are Ohdaira's FIR filter bank, Patterson's Gamma-tone filter bank, and Sneff's IIR filter bank, are examined their amplitude and phase characteristics as an inverse filter. We found that the Ohdaira's filter outperformed over other two filters. we could demonstrate a good quality reproduced speech with the reverse processing of auditory firing pattern processed with the forward auditory model. The processing are programmed in MATLAB and Mathematica. The results are going to present at the TCPhS'99 conference this summer.We improved an algorithm for conversion from speech signal into electrode pulse of the multi-channel cochlear implant based on an efficient speech coding technology. This problem is analyzed as a compressed speech information where auditory filters are switched one by one. This process is reduced as to find an optimal combination of one by one stimulation sequence of bipolar pulse to a given set of filter. Coding quality was perceptually evaluated with resynthesis of coded speech. High frequency components, that are relatively lower than low frequency component, are not well encoded and heard almost missing. This result was presented at the international conference in Australia ICSLP'98 in Sydney. We could have an interesting discussion with Australian progressed cochlear implant researchers including a famous inventor of the multi-channel cochlear implant system G.M.Clark. They were very much impressed in our approach and asked me many questions and proposed further cooperation. During this project term, our cochlear implant study is internationally recognized and marked an important contribution to the history of cochlear implant study.