Analysis and classification of hybrid EEG features based on the depth DRDS videos

Stereo vision cognition is a crucial advanced function of human beings, and stereoscopic acuity is an important index to detect stereo vision. Electroencephalograph (EEG) is an effective method of detection. Therefore, it has great significance to research the relationship between stereoscopic acuity and EEG signals for the development of 3D technology.

This paper proposes a multi-channel selection sparse time window common spatial group (MCS-STWCSG) multi-classification method. Firstly, a channel selection method based on improved common spatial pattern- (CSP-) rank is applied to select optimal channels to reduce redundant signal. Secondly, based on the one vs. one (OVO) computational model, we extend traditional CSP to the common spatial group (CSG) to implement three-classification. Finally, this paper optimizes time-frequency characteristics and hybrid signal features by sparse regression and utilizes a support vector machine (SVM) with radial basis function (RBF) kernel to identify depth dynamic random dot stereogram (DRDS) video tasks.

The selected channels are all located in and near the occipital region and time-frequency characteristics can acquire better classification results compared with frequency characteristics. The highest classification result can reach 94.67%.

The MCS-STWCSG multi-classification method optimizes features from multiple aspects and its performance is obviously better than other methods for hybrid EEG signals of depth DRDS.

Channel selection and time-frequency segmentation for feature extraction and classification algorithm of EEG signals can increase the classification accuracy. It proves the feasibility and accuracy of the proposed method.