On the Efficient Speech Feature Extraction Based on Independent Component Analysis
Neural Processing Letters
Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
| Hi-index | 0.00 |
The spatio-temporal EEG features are extracted by a two-stage ICAs. First, a spatial ICA is performed to extract spatially-distributed sources, and the second ICA is introduced in temporal domain for the coefficients of spatial sources. This 2-stage method provides much better features than spatial ICA only, and is computationally more efficient than single-stage spatio-temporal ICA. Among the extracted spatio-temporal features critical features are selected for the given tasks based on Fisher criterion. The extracted features may be applicable to the classification of single-trial EEG signals.