针对现有方法在眼电伪迹自动去除中存在有用信息丢失、伪迹分量识别困难的问题,提出了一种结合粒子群优化算法、独立成分分析和小波变换的伪迹自适应去除算法。首先,采用均方根误差和Pearson相关系数设计了粒子群优化算法的适应度函数,...针对现有方法在眼电伪迹自动去除中存在有用信息丢失、伪迹分量识别困难的问题,提出了一种结合粒子群优化算法、独立成分分析和小波变换的伪迹自适应去除算法。首先,采用均方根误差和Pearson相关系数设计了粒子群优化算法的适应度函数,利用优化算法实现了两个样本熵阈值的自适应设置;然后利用快速独立成分分析算法将脑电信号分解为统计独立分量,根据第一个样本熵阈值自动识别含伪迹分量,含伪迹分量经过4层小波分解得到5个小波分量,根据第二个样本熵阈值自动识别伪迹分量,将识别的伪迹分量置零;最后经过小波重构和逆变换,获得去除眼电伪迹的脑电信号。采用Graz data set A数据集进行实验验证,结果表明提出的方法能够实现多通道脑电信号伪迹的自动去除;采用Klados数据集进行实验验证,结果表明,与SE-CEEMDAN方法相比,采用提出方法实验获得的均方根误差降低了4.816,约38.2%,Pearson相关系数提高了0.025,约2.97%。展开更多
The characteristics of quadratic spline wavelet in noise reduction are analyzed. 1-D and 2-D equivalent filters of wavelet transform are presented. For proving its effectiveness, in the last section we usewavelet filt...The characteristics of quadratic spline wavelet in noise reduction are analyzed. 1-D and 2-D equivalent filters of wavelet transform are presented. For proving its effectiveness, in the last section we usewavelet filter bank to reduce the noise on the saccadic EOG successfully.展开更多
文摘针对现有方法在眼电伪迹自动去除中存在有用信息丢失、伪迹分量识别困难的问题,提出了一种结合粒子群优化算法、独立成分分析和小波变换的伪迹自适应去除算法。首先,采用均方根误差和Pearson相关系数设计了粒子群优化算法的适应度函数,利用优化算法实现了两个样本熵阈值的自适应设置;然后利用快速独立成分分析算法将脑电信号分解为统计独立分量,根据第一个样本熵阈值自动识别含伪迹分量,含伪迹分量经过4层小波分解得到5个小波分量,根据第二个样本熵阈值自动识别伪迹分量,将识别的伪迹分量置零;最后经过小波重构和逆变换,获得去除眼电伪迹的脑电信号。采用Graz data set A数据集进行实验验证,结果表明提出的方法能够实现多通道脑电信号伪迹的自动去除;采用Klados数据集进行实验验证,结果表明,与SE-CEEMDAN方法相比,采用提出方法实验获得的均方根误差降低了4.816,约38.2%,Pearson相关系数提高了0.025,约2.97%。
文摘The characteristics of quadratic spline wavelet in noise reduction are analyzed. 1-D and 2-D equivalent filters of wavelet transform are presented. For proving its effectiveness, in the last section we usewavelet filter bank to reduce the noise on the saccadic EOG successfully.