保留时域精细结构的高生物拟真全植入式神经形态人工耳蜗芯片设计

项目名称： 保留时域精细结构的高生物拟真全植入式神经形态人工耳蜗芯片设计

项目编号： No.61501445

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 王师惟

作者单位： 中国科学院深圳先进技术研究院

项目金额： 23万元

中文摘要： 本项目利用超大规模神经形态集成电路技术，解决未来人工耳蜗产品的多项技术难题。有别于传统的基于低阶带通滤波加连续交错采样的人工耳蜗系统架构，本项目采用高度匹配耳蜗基底膜高品质因数和高阻带衰减斜率带通特性的仿生耳蜗滤波器阵列以及模拟耳蜗外毛细胞功能的自适应滤波器控制模块实现声音信号的分频及压缩处理，采用基于异步增量调制的脉冲编码方式生成耳蜗神经刺激信号。上述策略旨在提高人工耳蜗声音信号处理机制的生物拟真并且使声音信号的时域精细结构在神经刺激信号中得以表达，从而提升人工耳蜗产品在噪声环境下的性能以及感知音乐等时频信息更为复杂的声音信号的能力。另外，本项目提出的人工耳蜗声音信号处理系统可以由低功耗模拟集成电路实现， 预计整体功耗可以控制在1毫瓦以下，从而满足全植入式人工耳蜗的要求。

中文关键词： 生物电子学；生物医疗电路与系统；人工耳蜗；超大规模神经形态集成电路；混合信号集成电路设计

英文摘要： This project aims to solve a number of technical challenges in future cochlea implant (CI) devices using neuromorphic VLSI techniques. Being different from the conventional CI system architecture that is based on low-order band-pass filters plus continued-interleaved-sampling, the proposed design in this project achieves the sound spectral segmentation and compression by respectively employing the biomimetic cochlea filters that highly match the high-Q and ultra-steep roll-off band-pass characteristics of the basilar membrane, and the self-adaptive control block that models the function of the outer hair cell in real cochlea. Such strategies aim to improve the bio-fidelity of the signal processing mechanism in CI devices and enable the representation of the temporal fine structure of sound input in the neural stimulation signals, by virtue of which the CI devices can have improved performance in noisy environment and be enabled to better perceive music and other sounds with more complex time-frequency information. Besides, the sound processing system is implementable using low-power analog VLSI circuitry, and its overall power consumption can be controlled within 1 milliwatt to satisfy the stringent power requirement of a fully-implantable CI device.

英文关键词： Bioelectronics;Biomedical Circuits and Systems;Cochlea Implant;Neuromoprhic VLSI;Mixed Signal IC Design