ABSTRACT

The artifacts such as ocular, muscle and certain electrical disturbances contaminate the Electroencephalogram (EEG). Wavelet enhanced independent component analysis (wICA) with static segmentation preserves little information of both spectral and coherent characteristics of neural activity. However, the considerable amount of valuable information can be preserved with novel automatic dynamic size segmentation of multichannel EEG signal with wICA. Most of the signal information lost in the threshold process (inappropriate threshold value) and this could be overcome with an adaptive threshold approach. It is assumed that the brain neural activity is a Gaussian random distribution with zero mean only for initial state.

The 16 channel EEG signals are acquired with Ocular Artifact (OA). The subject is instructed to blink an eye during the time of recording. A National Instrument (NI) data acquisition card is used to acquire the EEG data in MATLAB with sampling rate of 1024 Hz. Statistical parameters like Standard deviation, mean power (PSD), root mean square error (RMSE) are used for analysis and comparison. The proposed dynamic segmentation method is better for suppression of ocular artifacts which preserves the brain neural activity as compared with static segmentation. The artifacts related with eye blinking are removed completely and successfully preserve spectral and coherent characteristics of EEG activity of interest. It is proved that automatic dynamic segmentation is a key tool. The presented method is best suitable for suppression of OA and estimation of brain neural activity.

摘要

诸如眼睛、肌肉和某些电干扰之类的伪影会污染脑电图 (EEG)。具有静态分割的小波增强独立分量分析 (wICA) 保留的神经活动的光谱和相干特征的信息很少。然而，使用 wICA 对多通道 EEG 信号进行新型自动动态大小分割可以保留大量有价值的信息。大多数信号信息在阈值过程中丢失（不适当的阈值），这可以通过自适应阈值方法来克服。假设大脑神经活动是一个高斯随机分布，仅初始状态的均值为零。

16 通道 EEG 信号是通过眼部伪影 (OA) 获得的。受试者被指示在记录期间眨眼。使用美国国家仪器 (NI) 数据采集卡在 MATLAB 中以 1024 Hz 的采样率采集脑电数据。标准偏差、平均功率 (PSD)、均方根误差 (RMSE) 等统计参数用于分析和比较。与静态分割相比，所提出的动态分割方法更适合抑制保留大脑神经活动的眼部伪影。与眨眼相关的伪影被完全去除，并成功地保留了感兴趣的脑电图活动的光谱和相干特征。事实证明，自动动态分割是一个关键工具。