Wang Hui, Song Aiguo, Li Bowei, Xu Baoguo, Li Yangming
School of Instrument Science and Engineering, Southeast University, Nanjing, Jiangsu, China.
Biorobotics Lab, Electrical Engineering Department, University of Washington, Seattle, WA, USA.
Technol Health Care. 2015;23 Suppl 2:S249-62. doi: 10.3233/THC-150960.
Study of imagination offers a perfect setting for study of a large variety of states of consciousness.
Here, we studied the characteristics of two electroencephalographic (EEG) patterns evoked by two different imaginary tasks and evaluated the binary classification performance.
Fifteen individuals (11 male and 4 female, age range of 22 to 33) participated in five sessions of 32-channel EEG recordings. Only by analyzing the subjects' output EEG signals from the central parieto-occipital region of PZ electrode, under the circumstances of consciousness of relaxation-meditation or tension-imagination, we carried out the experiment of feature extraction for spontaneous EEG, as the subjects were blindfolded but asked to open their eyes all the same. The Hilbert-Huang Transform (HHT) was utilized to obtain the Hilbert time-frequency amplitude spectrum, and then with the feature vector set extracted, a two-class Fisher linear discriminant analysis classifier was trained for classification of data epochs of those two tasks.
The overall result was that about 90% (± 5%) of the epochs could be correctly classified to their originating task.
This study not only brings new opportunities for consciousness studies, but also provides a new classification paradigm for achieving control of robots based on the brain-computer interface (BCI).
对想象的研究为研究多种意识状态提供了一个理想的环境。
在此，我们研究了两种不同想象任务诱发的两种脑电图（EEG）模式的特征，并评估了二元分类性能。
15名个体（11名男性和4名女性，年龄范围为22至33岁）参加了5次32通道脑电图记录实验。在受试者处于放松冥想或紧张想象意识状态下，仅通过分析来自PZ电极中央顶枕区域的输出脑电图信号，让受试者蒙上眼睛但仍睁开眼睛，我们对自发脑电图进行了特征提取实验。利用希尔伯特-黄变换（HHT）获得希尔伯特时频幅度谱，然后提取特征向量集，训练一个两类Fisher线性判别分析分类器对这两项任务的数据片段进行分类。
总体结果是，约90%（±5%）的数据片段能够被正确分类到其原始任务。
本研究不仅为意识研究带来了新机遇，还为基于脑机接口（BCI）实现机器人控制提供了一种新的分类范式。
