Electroencephalogram (EEG) activities in the frontal area are profoundly influenced by non-cerebral activities such as electrooculogram (EOG) signals. In this work, we suggest the Least Mean Square-Fourth (LMS/F) adaptive method as a noise canceler to get rid of eye artifacts from raw EEG data. In this case, we make separate recordings of the reference signals such as the horizontal EOG and the vertical EOG. The signals that were captured are then processed using a FIR filter, the coefficients of which are adaptively updated using an LMS/F algorithm. In order to generate artifact free EEG signals, these signals are first removed from the original EEG data. When compared to traditional approaches such as Least Mean Fourth and combined step size normalized least Mean Square adaptive algorithm, the results of the newly suggested algorithm are superior in a variety of respects for both the synthetic data and the experimental data.

中文翻译：

---

用于从 EEG 信号中去除眼部伪影的最小均方/第四自适应算法

额叶区的脑电图 (EEG) 活动深受非大脑活动（例如眼电图 (EOG) 信号）的影响。在这项工作中，我们建议使用最小均方四（LMS/F）自适应方法作为噪声消除器，以消除原始脑电图数据中的眼部伪影。在这种情况下，我们分别记录水平EOG和垂直EOG等参考信号。然后使用 FIR 滤波器处理捕获的信号，并使用 LMS/F 算法自适应更新其系数。为了生成无伪影的脑电图信号，首先从原始脑电图数据中去除这些信号。与最小均值四分之一和组合步长归一化最小均方自适应算法等传统方法相比，新提出的算法的结果在合成数据和实验数据的各个方面都优于。