用于痫样脑电在线检测的gm-C小波滤波器实现理论与方法研究

项目名称： 用于痫样脑电在线检测的gm-C小波滤波器实现理论与方法研究

项目编号： No.61504008

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： 赵文山

作者单位： 北京交通大学

项目金额： 20万元

中文摘要： 穿戴式动态脑电（WAEEG）是癫痫诊断技术的前沿研究方向。目前WAEEG研究的焦点为脑电信号的在线数据缩减，其核心任务是痫样波形（EW）检测电路的设计。本项目拟研究EW检测用极低频gm-C小波滤波器的实现理论与方法，包括：(1)提出基于移动最小二乘的实小波基逼近模型和复小波基相位逼近模型，并利用非均匀采样方法优化求解，从而提高EW检测的精度和实时性。(2)提出基于矩阵标定理论的gm-C实小波滤波器优化结构，以多回路反馈结构为框架，利用矩阵运算优化节点方程系数矩阵，从而实现通用性和动态范围的协同提升；在此基础上，提出“共享型”gm-C复小波滤波器的优化结构，以降低系统的体积与功耗。(3)提出基于串联-并联电流镜技术的极低跨导值gm单元设计方法，以降低电路的噪声和失配误差。(4)利用真实脑电数据评估gm-C实小波和复小波滤波器的EW在线检测性能，完成0.18微米CMOS工艺下的流片和测试验证。

中文关键词： 穿戴式动态脑电；在线数据缩减；痫样波形检测；模拟小波滤波器；运算跨导-电容电路

英文摘要： Wearable ambulatory EEG (WAEEG) system has been considered as the frontier research in the field of epilepsy diagnosis. Currently, the research of WAEEG is focused on the online EEG data reduction, whose core task is to design the epileptiform waveform (EW) detection circuit. Under this background, this project plans to conduct research on the realization theory and method of extremely low-frequency gm-C wavelet filter for EW detection, mainly involving: (1) Propose a novel approximation method to enhance EW detection precision and facilitate real-time operation by constructing the approximation model for real-valued wavelet bases and the phase approximation model for complex wavelet bases on basis of Moving Least Square, and obtaining the optimization solution by utilizing non-uniform sampling strategy. (2) Propose a novel construction method for optimal gm-C real-valued wavelet filter structure based on matrix scale theory to collaboratively enhance the generality and dynamic range by selecting multiple loop feedback structure as basic frame and optimizing the coefficient matrix of nodal equation using matrix operations. Then, based on the aforementioned research, the construction method for optimal ‘structure-shared’ gm-C complex wavelet filter structure is presented to minimize the chip size and power dissipation. (3) Propose a design method for the transconductor cell with ultra-low transconductance by using series-parallel current mirrors, by which the noise and mismatch errors can be reduced. (4) Compare the EW online detection performance between the gm-C real-valued and complex wavelet filters by using real EEG data, and complete the chip fabrication and measurement of the selected gm-C wavelet filter by using 0.18um CMOS technology.

英文关键词： Wearable AEEG;Online Data Reduction;Epileptiform Waveform Detection;Analog Wavelet Filter;gm-C Circuit