LIN, CHIN TENG,林进灯,林進燈,YU-RUI MING,明玉瑞,明玉瑞,KO, LI WEI,柯立伟,柯立偉,HUANG, CHIH SHENG,黄志胜,黃志勝
申请号：
TW104126845
公开号：
TW201707648A
申请日:
2015.08.18
申请国别(地区):
TW
年份:
2017
代理人:
摘要：
The prevent invention provides a method for instantly removing artifacts of electroencephalogram, which using an electroencephalogram detection device to collect a users electroencephalogram signals, and the use of a processor to remove artifacts in the EEG signals. First, the present invention processes EEG signals to a windows mode, then uses a canonical correlation analysis for converting the EEG signals into canonical correlation components signals. The feature extraction is further used to extract component characteristics from canonical correlation components. Based on component characteristics, a Gaussian mixture model is adopted to estimate the probabilities from electroencephalogram components and artifacts components for each canonical correlation components. The probabilities of electroencephalogram components of all canonical correlation components are considered to reduce and restructure electroencephalogram signals without artifacts.本發明提供一種即時腦波去除雜訊之方法,其係利用一腦波檢測裝置收集一使用者之複數腦波訊號,並利用一處理器去除該腦波訊號中之雜訊。首先,本發明將腦波訊號進行視窗化處理,利用一典型相關分析法(Canonical Correlation Analysis)將腦波訊號轉換成典型相關成分訊號;從典型相關成分訊號中擷取出複數成分特徵,並依據該等成分特徵、利用一高斯混合模型計算出每一該典型相關成分訊號的腦波成分機率及雜訊成分機率;利用腦波成分機率還原得到去除雜訊後之腦波。
