TY - JOUR
T1 - Feature Extraction of Time-Amplitude-Frequency Analysis for Classifying Single EEG
AU - Yonghui Fang & Xufei Zheng
JO - Journal of Fiber Bioengineering and Informatics
VL - 2
SP - 261
EP - 271
PY - 2014
DA - 2014/07
SN - 7
DO - http://doi.org/10.3993/jfbi06201412
UR - https://global-sci.org/intro/article_detail/jfbi/4783.html
KW - Brain-computer Interfaces (BCI)
KW - Electroencephalogram (EEG)
KW - Motor Imagery (MI)
KW - Discrete Wavelet Transform (DWT)
KW - Amplitude-frequency Analysis (AFA)
AB - Feature extraction and feature classification are two important stages in most EEG based Brain-computer
Interfaces (BCI). The features extracted by DiscreteWavelet Transform (DWT) have a great relationship
with sampling frequency. On the other hand, the features extracted by Amplitude-frequency Analysis
(AFA) always ignore time information. In this paper, we proposed a feature extraction scheme based
on Time-Amplitude-Frequency analysis (TAF) for classifying left/right hand imagery movement tasks.
The time and frequency information are included in the proposed features. The Graz datasets used in
BCI Competition 2003 and the datasets collected in the lab of Electromagnetic Theory and Artificial
Intelligence of Chongqing University are used to show the effectiveness of the proposed features. The
simulation showed that the proposed features improved the classifying accuracy and the Mutual
Information (MI) for both datasets. The mutual information of TAF for Graz2003 dataset is 0.58
which is better than that of AFA and DWT.