EEG Signal Classification Using Manifold Learning and Matrix-Variate Gaussian Model

Table 1

The pseudocode for training the B2DDLPP feature extractor.

---

Algorithm: B2DDLPP

---

Inputs:

- Training sample . The total number of samples is N. The number of training samples in each class is , 1≤ i ≤ Z.

Outputs:

- The feature extraction operators and - the corresponding and values which determine the priority in selecting the elements in feature matrix.

Procedure:

1. Calculate the spatial covariance matrix and the spectral covariance matrix according to (18) and (19).

2. Calculate and according to (21) and (22).

3. Calculate the eigenvalues and the corresponding eigenvectors of , 1≤ ≤. And calculate the eigenvalues and the corresponding eigenvectors of 4. Construct U and V.

5. Calculate the feature matrix Y according to (24).

6. Choose the elements of Y which correspond to the d largest values. D is the dimension after feature extraction.

---