Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
General Tensor Discriminant Analysis and Gabor Features for Gait Recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
| Hi-index | 0.00 |
Brain Computer Interface (BCI) is a system that provides a nonmuscular communication between men and machines. This paper aims at classification of motor (hand movement) imagery to facilitate control for physically challenged persons using EEG signals. The work involves a scheme based on tensors. Advantages of this scheme over conventional schemes like Common Spatial Patterns (CSP), Linear Discriminant Analysis (LDA) are that 1. The number of parameters required to model the data is reduced, 2. This scheme works well without pre-processing (filtering, artifact removal etc.) of EEG signals, 3. Undersampling problem (number of training samples is less than dimension of data) is reduced. The work employs wavelet transform for representing EEG signals as tensors, General Tensor Discriminant Analysis (GTDA) for dimensionality reduction and Support Vector Machines for classification. Applications to datasets show the efficiency of this scheme compared to CSP and LDA. The work is expected to open new and higher levels of control for BCI since preprocessing is not needed.